Indicate (X) client(s) to whom this final report is submitted. Replace any of these with other relevant clients if required. FINAL REPORT FOR 2009

CFPA Canning Fruit Producers Assoc. Submit to: Wiehahn Victor PO Box 426 Paarl, 7620 Tel: +27 (0)21 872 1501 DFPT Deciduous Fruit Producers Trust Submit to: Louise Liebenberg Suite 275, Postnet X5061 Stellenbosch, 7599 Tel: +27 (0)21 882 8470/1 DFTS Dried Fruit Technical Services Submit to: Dappie Smit PO Box 426 Paarl, 7620 Tel: +27 (0)21 872 1501 Winetech Submit to: Jan Booysen PO Box 528 Paarl, 7624 Tel: +27 (0)21 807 3324 inmaak@mweb.co.za louise@dfptresearch.co.za dappies@dtd.co.za booysenj@winetech.co.za X Indicate (X) client(s) to whom this final report is submitted. Replace any of these with other relevant clients if required. FINAL REPORT FOR 2009 PROGRAMME & PROJECT LEADER INFORMATION Programme leader Project leader Title, initials, surname Mr. A.R. Mulidzi Dr. W. J. Conradie Present position Research Team Leader Specialist Scientist Address ARC Infruitec-Nietvoorbij Private Bag X 5026 Stellenbosch 7599 ARC Infruitec-Nietvoorbij Private Bag X 5026 Stellenbosch 7599 Tel. / Cell no. (021) 809 3070 (021) 809 3025 Fax (021) 809 3260 (021) 809 3260 E-mail mulidzir@arc.agric.za conradiek@arc.agric.za PROJECT INFORMATION Project number Project title WW13/01 Quantification of the effects of soil and climate on wine quality and character, aimed at scientific demarcation. Industry programme Fruit kind(s) CFPA DFPT DFTS Winetech Other Wine grapes Soil Science Start date (dd/mm/yyyy) 01/04/1993 End date (dd/mm/yyyy) 31/03/2005

2 FINAL SUMMARY OF RESEARCH PROJECT PROGRAMME & PROJECT LEADER INFORMATION Programme leader Project leader Title, initials, surname Mr. A.R. Mulidzi Dr. W. J. Conradie Institution ARC Infruitec-Nietvoorbij ARC Infruitec-Nietvoorbij Tel. / Cell no. (021) 809 3070 (021) 809 3025 E-mail mulidzir@arc.agric.za conradiek@arc.agric.za PROJECT INFORMATION Project number WW 13/01 Project title Fruit kind(s) Quantification of the effects of soil and climate on wine quality and character, aimed at scientific demarcation. Wine grapes Start date (dd/mm/yyyy) 01/04/1993 End date (dd/mm/yyyy) 31/03/2005 (Give a summary of the total project in no more than 250 words - Compulsory). A study in five commercial, rain-fed Sauvignon blanc vineyards, grown at different localities in the Western Cape of South Africa, was carried out over a period of 10 years. These localities (four within the district of Stellenbosch and one in Durbanville) were within a radius of 15 km and underlain by different geological formations. Two experimental plots, representing different soil forms, were identified at each locality, while a weather station was also erected at each locality. Despite their geographic proximity, meso-climates differed between the five localities, mainly on account of their various landscape positions and distance from the ocean. Maximum temperature for February differed by 3.1 C between the warmest and coolest locality. Ripening was also affected by soil type, resulting in differences of up to 12 days for time of harvest for grapes from different soil types, at the same locality. Climatic conditions also varied appreciably over the course of the investigation period, with 2001/2002 being the warmest (Winkler index = 1969), and 2002/2003 the coolest (Winkler index = 1717). These variations in seasonal climatic conditions affected soil water contents to a considerable extent. Experimental wines, prepared separately for each experimental plot, were evaluated annually by a trained panel. Overall wine quality was highest in 1997/1998 (mild water stress up to fruit-set, followed by good rain during November) and in 1999/1999 (low temperatures throughout the spring period). Quality was lowest when exceptionally high temperatures were experienced either during summer (1999/2000) or during spring (2001/2002). In general,

3 overall wine quality tended to be superior at the cooler localities (Helshoogte, Durbanville and Kuils River). In some cases (Helshoogte) wine style was largely affected by soil type, but overall wine quality remained high, irrespective of soil type. In other cases (Durbanville and Kuils River) wine quality tended to be reduced by lower soil water contents. At Devon Valley, wine quality was also lower for the soil with a lower water holding capacity. At Papegaaiberg wine quality was low, irrespective of soil type. Data from this project can be used to identify terroirs that will be able to produce Sauvignon blanc wine with a specific character.

4 FINAL REPORT (Completion of points 1-4 is compulsory) 1. Problem identification and objectives State the problem being addressed and the ultimate aim of the project. A viticultural terroir can be defined as a unit of the earth s biosphere that is characterized by relatively homogenous topographical, pedological and climatic features (Laville, 1993). With the aid of various management decisions these features will be expressed in the final product, resulting in distinctive wines with an identifiable origin (Carey, 2001). However, before viticultural terroirs can be identified correctly, sufficient information regarding the most important terroir parameters (climate, soil and cultivar), is essential. On account of South Africa being a relatively young wine producing country, such information is not readily available, resulting in vineyards currently being demarcated by technical experts on a theoretical basis (Saayman, 1998). Areas are allowed to express their specific wine style and character after demarcation, instead of proving originality beforehand. Demarcation of this nature is not specific enough to meet the demands of discerning consumers (Van Zyl, 2000), thus stressing the importance of the terroir approach (Carey et al., 2001; Conradie et al., 2002), which has been acknowledged as an important factor in European vineyards (Seguin, 1986; Falcetti, 1994; Morlat, 1996). In view of above-mentioned scarcity of scientific data concerning the demarcation of terroir units in South Africa, a study was undertaken at five selected localities, representing different landscape positions and underlain by contrasting rock types, in the Stellenbosch/Durbanville districts. These districts contain 17% of South Africa s vineyards (SAWIS, 2006), while Sauvignon blanc, being the most important cultivar for production of high quality white wines, was used as test material. Two different soil types were identified at each locality, resulting in 10 experimental sites. The hypothesis was that soil type may play an important role in determining wine characteristics, resulting in different wine styles for different soils, even under identical meso-climatic conditions. The ultimate goal of the project was to identify sets of conditions (climate, landscape position, altitude, soil, etc.) that will lead to wines with specific styles. The first objective was to quantify differences in meso-climate at the selected localities, to determine root distribution, as well as the physical and chemical characteristics of the experimental soils, while soil and plant water status were to be monitored on a regular basis. The interaction between climate, landscape position, geology and soil characteristics, as manifested in grapevine performance, was to be determined with the scope limited to phenology, growth, yield and chemical composition of grapes. If localities and soils can be separated in terms of above-mentioned parameters, it should be possible to produce, within the Stellenbosch/Durbanville districts, a range of Sauvignon blanc

5 wines with region-specific characteristics. Differences in actual wine quality will then become the subject of further investigation.. As a second objective seasonal changes in wine character were to be quantified, using the mean value from 10 experimental vineyards (5 localities x 2 soils) as point of departure. The mean seasonal value would be regarded a realistic guideline towards the quality/style of Sauvignon blanc, for that specific vintage. Differences between individual vintages would then be discussed in the light of seasonal changes in environmental factors, especially climate/soil interactions. As a third objective the extent up to which wine style differed from season to season at each of the 10 individual sites, were to be reported. This should make it possible to identify specific terroirs that will be able to produce Sauvignon blanc wine with a specific style during most seasons, irrespective of changes in environmental conditions thus attaining the ultimate goal. 2. Workplan (materials & methods) List trial sites, treatments, experimental layout and statistical detail, sampling detail, cold storage and examination stages and parameters. The study was carried out over 10 seasons (1994/1995 to 2003/2004) in five commercial, rain-fed Sauvignon blanc vineyards, at different localities in the Western Cape of South Africa. Experiment layout and experimental procedures have been described (Conradie et al., 2002). Briefly, two contrasting soil types were identified within each vineyard, using the South African Soil Classification System (Soil Classification Working Group, 1991). Experimental plots were selected on each soil type. Automatic weather stations (MC Systems, Cape Town), erected halfway between the two plots at each locality, recorded temperature, rainfall, radiation, hours of sunshine, wind speed and direction every minute. Details concerning co-ordinates, altitudes, soil types, climatic parameters and phenology, as described in a previous papers (Conradie et al., 2002; Conradie & Bonnardot, 2004; Conradie & Olivier, 2004), are summarized in Table 1. Experimental procedures entailed the following: Leaf water potential Leaf water potentials (pressure chamber technique of Scholander et al., (1965)) were measured once per week from November to March. Uncovered, fully mature sunlit leaves were used. Measurements were made on three leaves per experimental plot between 12:00 and 15:00.

6 Soil water Changes in soil water content were measured weekly at 300 mm depth intervals, down to a depth of 1200 mm, using the neutron scattering technique. Neutron counts were calibrated against gravimetric soil water contents (mass %), starting in August ( field capacity) and ending in March ( permanent wilting point). Soil bulk density values (determined on undisturbed soil cores) were used to convert soil water content (mass %) to volumetric soil water content (θv) for each layer (0-300 mm, 300-600 mm, 600-900 mm and 900-1200 mm). Best fit for the relationship between neutron probe measurements and θv was obtained with linear equations (Equation A). Water holding capacities (.0025 MPa to 0.4 MPa) were determined on undisturbed soil cores, sampled according to the different horizons, using standard pressure plate equipment. The best fit between soil matric potential (Ψ m ) and θv, for the different soil depths, were obtained with power equations (Equation B). Equations A and B were combined in order to obtain a direct relationship between neutron probe measurements and Ψ m. Experimental wines Winemaking: Grapes were harvested at optimal ripeness, which was considered to be at approximately 23 B, at a titratable acidity of 8 g/l and a ph of between 3.0 and 3.2. Forty to 60 kg of grapes was harvested from each plot. After being crushed and de-stemmed, grapes underwent a skin contact period of six hours at 14 C. Apart from 50 mg/l of di-ammonium phosphate being added before fermentation, experimental wines were prepared and bottled as described by Conradie (2001). Two replicates were fermented separately for each plot. This meant that 20 wines were prepared annually (five localities x two soils x two replicates). Sensorial evaluation: Wines were evaluated two months after bottling, i.e. six months after fermentation, by a panel of 14 trained wine tasters. Tasters were trained in preliminary sessions on the evaluation of Sauvignon blanc wine, e.g. odour recognition. Tasting took place in individual booths under white light. Wines were served in completely randomized order. A ten-centimetre unstructured line scale was used and the judges were asked to mark wines from undetectable to prominent for specific aromatic parameters. Based on the standardized system of wine aroma terminology (Noble et al., 1987), the latter entailed aroma intensity, fresh vegetative character (bell pepper, freshly cut grass, eucalyptus and mint), cooked vegetative character (green beans, asparagus, olive, artichoke), dry vegetative character (hay/straw, tea, tobacco), spice character (liquorice, aniseed, black pepper, clove), tropical fruit (pineapple, melon, banana, guava) and dried fruit (strawberry jam, raisin, prune, fig). Overall wine quality and fullness were also rated on an unstructured line scale, from unacceptable to excellent.

7 Statistical analyses Climatic data were analyzed statistically, using the 10 seasons as replicates. The sensory attributes were subjected to a factorial analysis of variance and the repeated measurements over the 10 seasons were used as a subplot factor. Student s t=lsd were calculated at a 5% significance level to compare means of significant effects. 3. Results and discussion State results obtained and list any benefits to the industry. Include a short discussion if applicable to your results. This final discussion must cover ALL accumulated results from the start of the project, but please limit it to essential information. First objective Aims: Quantification of differences in meso-climate, physical and chemical characteristics of the experimental soils, root distribution, grapevine performance as indicated by phenology, growth, yield and chemical composition of grapes. Results, as already described (Conradie et al., 2002), can be summarized as follows: Climate: The study region, located in the winter rainfall area at the extreme south western part of the Western Cape Province, receives the bulk of its rainfall from late autumn to the beginning of spring. Data for the 1994-2003 period (Table 2), indicated that 70% (483 mm) of the annual total (687 mm) was recorded from March to August (autumn and winter months), 21% (142 mm) during spring (September to November) and only 9% (62 mm) during the summer months (December to February). The dry and warm summers are similar to those experienced in the Mediterranean region. During the study period, maximum temperature averaged 27.7 C during summer, while the mean temperature of the warmest month (February) was recorded as 21.6 C. The study region is suitable for the production of high quality red and white table wine (De Villiers et al., 1996). According to the Winkler degreeday classification (Winkler et al., 1974), the study region (index value = 1810) can be placed in Winkler s Region III. Despite their geographic proximity, significant differences in finer climatic aspects could be detected between the different localities, especially during summer. This was mainly due to their various landscape positions and distance from the ocean. Number of hours (Dec-Feb) with temperatures >30ºC, growing degree days (Dec-Feb) and February maximum temperature pointed towards Kuils River, Helshoogte and Durbanville as being cooler than Papegaaiberg and Devon Valley. Maximum temperature for February differed by 3.1 C between the warmest (Papegaaiberg) and coolest (Durbanville) locality. The lowest

8 temperature variability index was experienced at the localities closest to the sea (Kuils River and Durbanville), due to moderate maximum and minimum temperatures. The inland stations of Papegaaiberg and Devon Valley experienced a more continental climate (highest temperature variability index). Helshoogte also experienced a more continental climate, but was cooler than Papegaaiberg and Devon Valley, due to its location at a higher altitude. This locality also experienced highest rainfall during the active growing season (Oct-Mar) and highest number of hours with temperatures below 12 C (due to cooler nights). Physical and chemical characteristics of the experimental soils: Particle size analyses suggested that four of the soils (Tukulu at Kuils River, Oakleaf at Devon Valley, both Westleigh and Tukulu at Durbanville) were related to underlying parent materials. The other six soils must have developed from admixtures of materials. This is a common phenomenon in the Western Cape. No major differences in chemical analysis occurred between soils at the same locality. Even though the concentrations of some elements (P, K and Ca) varied appreciably between localities, grapevine performance should not have been be unduly impeded, either on account of deficiencies or toxicities, at any of the localities. The highest C-content was found for Helshoogte and Durbanville, suggesting that the soils at these localities may have been formed under cooler conditions (Stevenson, 1986). The lowest value for K in the subsoil was found at Durbanville, where the soils originated from phyllitic shale, thus being in agreement with the fact (Visser, 1964; Wooldridge, 1988) that shales contain less total potassium than granite. At the other localities, however, potassium levels of subsoils could not be related to underlying geological formations, possibly because management practices, such as liming and fertilization, changed the chemical properties of the soils. Root distribution: Root distribution was mostly affected by factors such as soil moisture, compacted layers and percentage stone, and not directly related to soil form. Furthermore, relatively good root distribution was obtained for at least one soil from each individual parent material (granite, hornfels, shale), suggesting that root distribution is only indirectly affected by parent material. Phenology, growth, yield and chemical composition of grapes: As already mentioned, significant differences in finer climatic aspects could be detected between the different localities, with Kuils River, Helshoogte and Durbanville being cooler than Papegaaiberg and Devon Valley (Table 1). This resulted in date of budburst differing by six days (10 th September to 16 th September) between the earliest and the latest site, while date of flowering differed by 10 days (1 st November to 11 th November). Harvesting at one of

9 the cooler localities (Kuils River), as well as the two warmer localities (Papegaaiberg and Devon Valley) occurred within three days (9 th February to 12 th February). Low water holding capacities for the soils at Kuils River resulted in relatively high water stress during the latter parts of the season, leading to earlier ripening, in spite of relatively cool conditions (Conradie et al., 2002). In view of grapes generally being harvested before middle February at these three localities, wine style should not have been affected by climatic conditions during the latter part of February. The other two cool localities (Helshoogte and Durbanville) were harvested one to three weeks later (19 th February to 3 rd March) and therefore climatic conditions during the latter part of February may have affected wine style for Sauvignon blanc. In some cases ripening was also affected by soil type, with grapes from the wetter soil at Durbanville (Westleigh) being harvested 12 days later than those from the drier soil (Tukulu). In general, cane masses were highest at the localities (Helshoogte and Durbanville) where soils contained relatively high amounts of organic C, usually in combination with a high soil water status. Yield was lowest for the Westleigh at Durbanville, probably on account of infertility, induced during spring by too vigorous growth at floral initiation and therefore lower sunlight interception in the vine canopy. In contrast, yield was also low for the Tukulu at the warmest locality (Papegaaiberg). In this case low yield could be ascribed to a combination of factors (high temperatures and soil with a high gravel fraction and a low water holding capacity/poor root distribution). Berries were largest at the coolest locality (Durbanville) and smallest at one of the warmer localities (Devon Valley). At three of the localities (Kuils River, Papegaaiberg and Helshoogte) berry size was also significantly affected by soil form. Acidity was lowest, and ph highest, for grapes from the warmest locality, but low acidity and moderately high ph values were also found for grapes from the coolest locality, probably on account of differences between clones. However, within individual localities, acidity tended to be lowest and/or ph highest where water stress was the highest. Conclusions: Within the Western Cape of South Africa appreciable climatic differences can occur between localities in close proximity. This is largely on account of different aspects, altitudes and distances from the sea. These subtle climatic differences do have an effect on phenology and on the chemical composition of Sauvignon blanc grapes. Ripening can be either enhanced or delayed, while sugar/acid/ph balances are also affected. These differences should be reflected in wine quality. In spite of climate appearing to be the most important driving force affecting grapevine performance, the effect of soil could not be discounted. Especially in the case of soils with

10 low water holding capacities the advantages of a good climate could not always be fully exploited. High water stress sometimes resulted in grapes with low acidity/high ph. On the other hand, adequate ripening was also problematical on soils that were too wet. It has been argued that application of the terroir concept is of doubtful value in relatively warm winegrowing countries, like South Africa. Results from this study, however, suggested that it should be possible to produce, within the Stellenbosch/Durbanville districts, a range of Sauvignon blanc wines with region-specific characteristics. Differences in wine style, as induced by different terroirs, should be the subject of further investigation. Second objective Aims: Seasonal changes in wine character were to be quantified, using the mean value from 10 experimental vineyards (5 localities x 2 soils) as point of departure. Differences between individual vintages would be explained at the hand of seasonal changes in environmental factors, especially climate/soil interactions. Results, as already described (Conradie & Bonnardot, 2011a), can be summarized as follows: Seasonal variations in soil water content: In general, readily available water (water retained between -0.01 MPa and -0.10 MPa) was depleted in the 0-300 mm soil layers by end November. Depending on locality and soil type, matric potentials of the 300-600 mm layers would usually drop below -0.10 MPa sometime during December. Seasonal variations in soil water contents of the deeper soil layers (600-900 mm and 900-1200 mm) will be discussed in more detail elsewhere (Conradie & Bonnardot, 2011b), but as an example the values for Helshoogte are shown in Figure 1. In the case of the Tukulu soil (Fig. 1a), matric potential in the 600-900 mm layer would usually drop below -0.10 MPa between middle December (pea size berries) and middle January (post véraison). In the case of the 900-1200 mm layer, matric potential did not decrease below -0.10 MPa before middle January, while values were still higher than -0.20 MPa at the end of March (approximately one month after harvest). Even though matric potentials in 600-900 mm layers differed to some extent between the Tukulu and the Hutton soil (Fig. 1c), values were of comparable magnitude. In the case of the 900-1200 mm layers, however, values for the Hutton tended to be lower than -0.20 MPa at the end of the season (Fig 1d). The latter suggested that grapevines on the Hutton soil could have been subjected to higher water stress, in comparison to those on the Tukulu soil.

11 At the other localities (data not shown) differences could also be detected between different soil types especially towards the end of the season. Average wine quality during the investigation period: On account of a slightly different scorecard being used during the first three seasons (1994/95 to 1996/97), and in view of the vineyard at Helshoogte being uprooted at the end of the 2002/03 season, this discussion will focus primarily on the six seasons from 1997/98 to 2002/03. Mean scores (5 localities x 2 soils) for wines obtained during this period are displayed in Table 3. Over the course of the investigation the scores for fullness and aroma intensity varied by 18%, from 4.87 to 5.82 and from 5.82 to 6.96, respectively. The score for overall wine quality ranged from a high of 5.96 to a low of 5.29, thus varying by 12% only. The relatively low average scores for fullness (5.31), aroma intensity (6.28) and overall quality (5.52) are the results of wines prepared for experimental purposes only, thus not being market ready. Consequently, above-mentioned scores are within the normal ranges for experimental Sauvignon blanc wine, suggesting that average wine quality was acceptable during all seasons. Fresh vegetative character was the most prominent aroma component (3.52), followed by tropical fruit (3.09) and cooked vegetative character (2.49), thus supporting the view that the most distinctive characteristics of Sauvignon blanc wine can be defined as green and fruity/tropical aromas (Marais et.al., 1999; Lund et al., 2009). The score for fresh vegetative character varied by 42%, from 4.32 (highest) to 2.84 (lowest), while tropical fruit characteristics appeared to have remained relatively constant, varying by 19% only from a high of 3.34 to a low of 2.74. Cooked vegetative character showed variation of approximately 100% (from a low of 1.12 to a high of 3.62). Dry vegetative (1.55) and spice (0.61) characteristics were less prominent but showed appreciable seasonal variations. In the rest of the discussion these mean values, obtained for the different sensorial parameters over a period of six years, will be accepted as intermediate levels for Sauvignon blanc from Stellenbosch/Durbanville. High, Low and Intermediate ranges were identified for each individual sensorial component, at the hand of LSD values (Table 3). Values in the high range differ significantly from those in the low range. In the next section wine style/quality, as obtained during individual seasons, will be discussed in relation to the environmental conditions (rain fall, temperature and soil water content) experienced during that specific season. The scores for fullness, aroma intensity and overall quality will be used to distinguish between superor, average and below average vintages. Most pertinent results, as obtained during individual seasons, are summarized in Table 4.

12 1997/1998: High temperatures at the start of the growing season (up to fruit-set) may have been the reason why none of the individual aromatic components were scored in the high range. However, on account of high values for overall quality and aroma intensity the 1998 vintage had to be classified as superior. Mild to moderate water stress, as experienced at the start of the season, may have been beneficial to wine quality (Dry et al., 2001). Relatively cool conditions during the middle part of the season (November to January), together with good rain in November could also have been positive factors. High temperatures during February must have resulted in suboptimal conditions during the latter part of the ripening period, especially at localities where grapes were harvested relatively late. In general, wines were highly regarded, in spite of individual aromatic components not being exceptionally prominent. 1998/1999: In comparison to the other seasons, the 1999 vintage obtained the highest scores for fullness, aroma intensity and overall quality, pointing towards an exceptionally superior vintage. The first part of the season (up to end December one week before veraison) may have been conducive to high wine quality (relatively high rain fall in November and December, while temperatures were relatively low up to the end of December). High temperatures and low soil water during the last part of the season (January to March) may have impacted negatively on wine quality where grapes are harvested after the middle of February. However, the quality of the 1999 vintage may well have been the highest obtained during the six-year-period. Wine style (prominent green aromas with a low fruity component) was typical for Sauvignon blanc from a cool climate. 1999/2000: The scores for fullness, aroma intensity and overall quality were the lowest encountered during the investigation period. Consequently, 2000 had to be classified as a below average vintage. Most individual aromatic components also received the lowest scores allocated during the study period. Extremely high temperatures during December and January (from pea size berry stage to approximately two weeks before harvest), accompanied by low soil water contents (Fig. 1a - 1d), must have contributed to low levels of aromatic compounds. High temperatures during the preceding winter (1999) may also have contributed towards the disappointing wine quality. Wines were fairly neutral, without exhibiting the fruity characteristics expected during a warm season. 2000/2001: Even though fullness was scored in the high range, aroma intensity and overall quality received intermediate scores only, suggesting that the 2001 vintage was of average quality. In view of the fact that three of the aroma components (cooked vegetative, tropical

13 fruit and spice) were scored in the high range, with the other two (fresh vegetative and dry vegetative) at intermediate levels, above-mentioned was surprising, especially in comparison to the 1998 vintage. With individual sensorial parameters in most cases more prominent in 2001 than in 1998, higher overall quality could have been expected in 2001. However, it should be borne in mind that the score for aroma intensity (and overall quality) depends on the panel s judgement of the typicity of the aroma and/or the balance between the different aroma components. As in 1999/2000, exceptionally high temperatures and low rain fall during the 2000/2001 winter may have been negative for wine quality. While mild water stress up to the end of October may have been beneficial to wine quality in 1997/98, water stress (largely on account of low rainfall), persisted up to the end of January in 2000/2001. 2001/2002: Fullness was scored at the intermediate level, while aroma intensity and overall quality was regarded as low. Consequently, 2001/2002 could be classified as a below average season, together with 2000 vintage. Exceptionally high spring temperatures (Table 2), highest experienced during the investigation period, must have had a negative effect on wine quality. The 2002 vintage scored highest for tropical fruit aroma (3.34) and lowest for fresh vegetative aroma (2.84), thus complying with the fruity style expected during a warm season. In comparison to the 2000 vintage (second warmest season, on account of high summer temperatures), fresh vegetative character was lower 2002, but the other aromatic parameters (cooked vegetative, dry vegetative, tropical fruit and spice) were lower in 2000. High rain fall during December should have had a positive effect for the 2002 vintage, in comparison to 1999/2000. 2002/2003: Fullness and overall quality were scored in the intermediate range, with aroma intensity in the low range. Consequently, the 2003 vintage had to be classified as being of average quality, together with the 2001 vintage. Climatic conditions during this season appeared to have been conducive for the formation of fresh vegetative, dry vegetative and spice characteristics, resulting in all of these being scored in the high range. Cooked vegetative aroma was enhanced to a lesser extent and was scored in the intermediate range. Similar to the 2001 vintage, tropical fruit aroma was still in the intermediate range, in spite of the cool season. In view of the good aroma profile it was surprising that the panel was not impressed with aroma intensity. Similar to the 2001 vintage, the individual aromatic components may have been unbalanced. The low score for overall wine quality may also have been on account of wines not having been completely settled at the time of evaluation (six months after harvest). For some of the wines higher scores were obtained when an evaluation was done 18 months after harvest (not shown). It should also be borne in mind that different responses were obtained at different localities and for different soil types. Only

14 the average trend, obtained for 10 soils, is depicted in Table 3. Responses for individual terroirs will be reported in a subsequent publication (Conradie & Bonnardot, 2011b).. Conclusions Differences in wine style could, to a large extent, be related to differences in environmental conditions (Table 4). On account of relatively low temperatures up to end December, accompanied by good rain in November and December, high quality could have been expected in 1998/1999. On the other hand, due to exceptionally high temperatures, either during spring or during summer, lower quality could have been forecasted in 1999/2000 and in 2001/2002. In some seasons, however, wine style/quality differed from what would (normally) have been expected. In 1997/1998 quality was high, in spite of high temperatures during September and October. Mild water stress up to fruit-set, accompanied by good rain in November, may have been positive factors. In contrast, higher quality could have been expected in 2000/2001 and in 2002/2003, when temperatures were relatively low. During both seasons low soil water contents, on account of rain fall having been below average during winter and/or spring, may have been detrimental to wine quality. All of abovementioned should be confirmed in further studies. Third objective Aims: the extent up to which wine style differed from season to season at each of the 10 individual sites, were to be characterized. This should make it possible to identify specific terroirs that will be able to produce Sauvignon blanc wine with a specific style during most seasons, irrespective of changes in environmental conditions thus attaining the ultimate goal. Differences between soil types: Soil forms, as indicated in Table 1, were identified at the hand of the South African Soil Classification System (Soil Classification Working Group, 1991). According to this system, soils are classified as a first category into soil forms at the hand of the presence and properties of master horizons. Further classification into families is based on A, B and E horizon properties, degree of leaching, clay movement and wetness. Fairly large differences may, therefore, occur within a soil form. Consequently, the four Tukulu soil forms in Table 1 differed appreciably, even though signs of wetness could be detected in the subsoil in all cases. From a viticultural viewpoint, the capability of a specific soil to produce wine of outstanding quality (under rain-fed conditions) will be largely determined by root distribution and water holding capacity, as summarized in Table 5. On

15 account of readily available water usually being depleted in the 0-300 mm and 300-600 mm horizons towards the end of December, the rest of this discussion will focus primarily on the 600-900 mm and 900-1200 mm horizons. Wine style/quality: comparison between different localities: Mean values, as obtained at the different localities from 1997/1998 to 2002/2003, are shown in Table 6. At the hand of the scores allocated for fullness, aroma intensity and overall quality, the different sites (terroirs) could be classified as follows: Highest (superior) wine quality: Both sites at Helshoogte (Tukulu and Hutton soil types), one site at Durbanville (Tukulu soil type) and one site at Kuils River (Vilafontes soil type). Average (intermediate) wine quality: One site at Durbanville (Westleigh) and one site at Devon Valley (Oakleaf soil type). Below average (lowest) wine quality: One site at Kuils River (sandy Tukulu soil type), one site at Devon Valley (Glenrosa soil type) and both sites at Papegaaiberg (Avalon and Tukulu soil types). This classification indicated that wine quality was highest at the coolest sites, i.e. Helshoogte (high altitude, continental climate with higher temperature variability index), Durbanville and Kuils River (lower altitude, with a maritime climate, due to the proximity of the ocean). However, at Durbanville and at Kuils River soil type plated a significant role as well, with wine quality being reduced for the Westleigh and Tukulu, respectively. At Durbanville wine quality still tended to above average for the Westleigh, but at Kuils River wine quality was reduced to the lowest range for the sandy Tukulu. At the second warmest locality (Devon Valley) soil type also played a role, with wine quality being higher for the Oakleaf than for the Glenrosa soil type. At the warmest locality (Papegaaiberg) wine quality was in the lowest range for both soil types. The extent up to which wine quality/wine style was affected by climate/soil type during the different experimental seasons will be discussed in the next section. Helshoogte: As indicated in Table 5, both soils were of high potential. Water holding capacities were similar, but rooting density was higher for the Tukulu. During wet seasons superior drainage may be advantageous for the Hutton, while the reverse may be true during warm and dry seasons.

16 Over the course of the investigation period (Table 6), wines from both soils received high scores for aroma intensity, overall quality and cooked vegetative character, while fullness, dry vegetative character and spice characteristics were scored in the intermediate range. Wine from the Tukulu scored higher for fresh vegetative character, while a reverse trend was observed for tropical fruit character. This suggested that the wine from the Tukulu exhibited more vegetative- and less fruity characteristics, in comparison to the one from the Hutton. However, vegetative characteristics were still prominent in the wine from the Hutton soil, with tropical fruit characteristics relatively low, in comparison to the wines from Kuils River and Devon Valley. Different styles for the two wines from Helshoogte may well have been on account of differences in the availability of soil water (Figs 1b & 1d). In general (with 2000/2001 being the only exception), soil matric potential never dropped below -0.20 MPa in the 900 1200 mm horizon of the Tukulu soil. For the Hutton soil, however, soil matric potential was generally lower than -0.20 MPa in the 900-1200 mm horizon towards the end of the season. Higher levels of water stress must have induced more prominent fruity characteristics in wine from the Hutton. Responses during individual seasons can be summarized as follows: Fullness, aroma intensity and overall quality: The scores for the 1998 vintage (Table 7) were in agreement with the long-term average in Table 6 (aroma intensity and overall quality in the highest range, with an intermediate value for fullness). In general, the 1999 vintage received high scores, thus being in agreement with the overall pattern identified for this vintage (Conradie & Bonnardot, 2011a). However, in contrast to a relatively low score for the average wine from 2000 vintage, the Helshoogte scores were still in the higher ranges for this vintage. It has been argued that exceptionally high temperatures during December and January may have been a detrimental factor at most localities. During December and January of this season (1999/00), soil matric potentials were also relatively low at Helshoogte, especially in the 600-900 mm horizon (Figs 2a & 2b). However, the fact that wines from Helshoogte were affected to a lesser extent, may have been on account of grapes from Helshoogte normally being harvested two to three weeks later than the other sites (Table 1). This suggested that the grapes at Helshoogte may have been in a different developmental stage when high temperatures were experienced, resulting in wine style being affected to a lesser extent. Differences in the time of ripening may also have been the reason for wine from the Hutton (overall quality = superior) being affected to a lesser extent, in comparison to the one from the Tukulu (overall quality = average, with fullness below average). The 2001 vintage scored in the highest range for fullness, aroma intensity and overall quality, thus pointing towards wines of superior quality. In contrast, the average wine from Stellenbosch/Durbanville received an intermediate score only, probably on

17 account of a very warm and dry winter, resulting in soils being marginally dry towards the end of the growing season at some localities (Conradie & Bonnardot, 2011b). The latter was obviously less of a problem at Helshoogte, probably on account of higher rainfall during spring and summer, in comparison to the other localities (Conradie et al., 2002). For the 2002 vintage, however, overall wine quality for Helshoogte (average for the two soils) was reduced by 11.0% (5.82 to 5.18), while the average value for Stellenbosch/Durbanville was reduced by 3.1% only (5.52 to 5.35). This suggested that the negative effect of environmental conditions, as experienced during 2001/2002, were more pronounced at Helshoogte than at the other localities. Exceptionally high temperatures during spring (Conradie & Bonnardot, 2011a), and/or relatively wet soils up to harvest, must have been more detrimental at Helshoogte than at the other localities. In 2002/2003, the Tukulu soil yielded wine of superior quality, while most sensorial parameters were scored in the intermediate range only for the wine from the Hutton. This may have been on account of soil matric potential in the subsoil (900-1200 mm) being appreciably lower for the Hutton than for the Tukulu towards the end of the season (Fig 2b). This phenomenon has been ascribed to low rainfall during winter and spring (Conradie & Bonnardot, 2011a). Fresh-, cooked- and dry vegetative characteristics: For both soils, fresh vegetative character was high for the 1999, 2001 and 2003 vintages. According to the Winkler index, these were the coolest seasons, largely on account of low temperatures during the spring period (Table 2). Fresh vegetative character was down to intermediate levels for the 1998 and 2000 vintages, thus coinciding with the 4 th and 5 th highest Winkler indexes. During the warmest season (2001/2002), with a Winkler index of 1969, fresh vegetative character was reduced to the lowest range for wine from the Hutton soil, while it was still in the intermediate range for wine from the Tukulu. Above-mentioned suggested that fresh vegetative characteristics were largely affected by temperature, even though the effect of soil could not be completely discounted. In the case of cooked vegetative character, high values were also obtained for the 2001 and 2003 vintages, but values were in the intermediate range only for the other cool season (1999 vintage). During the latter season, spring temperatures were exceptionally low (GDD = 478), while summer-temperatures (GDD = 999) were high, being surpassed by the 1999/2000 season only (GDD = 1055). This suggested that cooked vegetative character may have been negatively affected by high summer temperatures (1999 and 2000 vintages), while high temperatures during spring (e.g. 1998 and 2002 vintages) had less effect. Dry vegetative character also seemed to be negatively affected by high summer temperatures (1999 and 2000 vintages), while highest values were recorded during the coolest season (2003 vintage). Above-mentioned suggested that fresh vegetative characteristics may have been directly correlated to the Winkler index, while this was not necessarily the case for cooked- and dry vegetative characteristics.

18 Tropical fruit- and spice characteristics: For Stellenbosch/Durbanville as a whole, tropical fruit was lowest during seasons with high summer temperatures (1999 and 2000 vintages). The trend for 2000 was similar at Helshoogte, but tropical fruit character was high for the 1999 vintage (cool spring, warm summer). Furthermore, values were also low for the 1998 vintage (relatively cool summer). This indicated that, in comparison to the other localities, vineyards at Helshoogte responded differently to environmental conditions experienced during 1997/1998 and 1998/1999. For the 2000, 2001, 2002 and 2003 vintages, tropical fruit character was clearly less pronounced in wine from the Tukulu soil, in comparison to the Hutton soil, thereby stressing the importance of soil/climate interactions. Spice character was lowest in the warmest seasons (1998/1999, 1999/2000 and 2001/2002), thus corresponding with the trend identified for fresh vegetative character. Summary: In general, the climate at Helshoogte appeared to be favourable for the production of wines of superior quality, while both experimental soils were of high potential. This resulted in wines of high quality, with 2001/2002 (high temperatures during the spring period and excessive rain during early summer) being the only exception. However, wine style differed between the two soil types. Wine from the Tukulu soil generally exhibited a strong fresh vegetative character, in combination with a cooked vegetative character. It may be postulated that it will be possible to produce wines of similar character, if terroirs with a similar climate and similar soil types can be identified. Durbanville: As indicated in Table 5, the Westleigh soil at Durbanville showed signs of a water table during winter. Probably on account of this, rooting density was much better for the Tukulu soil. Grapevines on the Westleigh tended to grow very vigorously during the early part of the season (Conradie et al., 2002), probably on account of the high water holding capacity. However, from November onwards soil matric potentials generally tended to be lower for the Westleigh than for the Tukulu (Figs 3a & 3b). Over the course of the investigation period (Table 6), wines from both soils received high scores for overall quality, with aroma intensity, cooked vegetative-, dry vegetative- and spice characteristics in the intermediate range. Fullness and fresh vegetative character were in the intermediate range for the Westleigh and high for the Tukulu. The wine from Westleigh received the highest score for tropical fruit character, even though this score was still relatively low in comparison to Kuils River and Devon Valley. Wine style was thus also affected by soil type, even though differences were not as prominent as those identified at Helshoogte. Responses for the different sensorial parameters, over the course of the investigation period, can be summarized as follows:

19 Fullness, aroma intensity and overall quality: The scores for the 1998 vintage (Table 8) were in agreement with the average values obtained over the experimental period (overall quality in the highest range, with aroma intensity and fullness generally in the intermediate range). Similar to Helshoogte, the 1999 vintage was in the superior range for the Tukulu, while quality was above average for the Westleigh. In 1999/2000, wine from the Westleigh received low scores for fullness and overall quality, while aroma intensity and overall quality were still high for wine from Tukulu, even though fullness was also in the low range. This pointed towards quality being below average for the Westleigh, while wine from the Tukulu could be classified as being of (at least) average quality. The inferior performance of grapevines on the Westleigh soil was probably on account of being subjected to higher water stress, especially towards the end of the season (Figs 3a & 3b). On 1 st March soil matric potential was down to approximately -0.25 MPa and -0.50 MPa in the 600-900 mm and 900-1200 mm horizons of the Westleigh soil, while values were still appreciably higher for the Tukulu (-0.07 MPa and -0.35MPa, respectively). Due to technological problems, experimental wines could not be prepared for the Tukulu 2000/2001 but, as in the case of Helshoogte, wine quality was superior for the Westleigh. In the 2001/2002 season, wine from the Tukulu scored in the superior range, while the Westleigh produced wine of average quality. As already mentioned, the high temperatures experienced during the spring 2001/2002, did not affect all localities in a similar way. In contrast to the situation at Helshoogte, high rain fall during December may have been beneficial at Durbanville. The 2003 vintage scored in the superior range for the Westleigh, but quality was below average for the Tukulu. This was probably on account of the 900-1200 mm horizon of the Tukulu being exceptionally dry from end January (Figure 3b). Soil matric potential in this horizon was also lower than in the shallower (600-900 mm) horizons (Fig 3a). A comparable situation has been reported for the Hutton at Helshoogte, and may be ascribed to insufficient rainfall during the preceding winter and spring. This meant that wines from both soils could be classified as below average for one vintage only (2000 for the Westleigh and 2003 for the Tukulu). Fresh-, cooked- and dry vegetative characteristics: As in the case of Helshoogte, fresh vegetative character was highest (both soils) in the seasons (1998/1999 and 2002/2003) with the lowest Winkler indices. Fresh vegetative character was also lower for the 1998 and 2002 vintages, when Winkler indices were high, largely on account of high spring temperatures (Table 2). Similar to fresh vegetative character, high values were obtained for cooked vegetative characteristics for the 1999 and 2003 vintages, while lower values, especially in the case of the Westleigh, were recorded for the 1998, 2000 and 2002 vintages. The latter suggested that cooked vegetative character may be negatively affected by high spring