1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
Vineyard variability in Marlborough, New Zealand: Characterising spatial and temporal changes
in fruit composition and juice quality in the vineyard
16
Previous work has demonstrated that vineyards are spatially variable and that this variability can be
17
understood in terms of the underlying characteristics of the land (soils, topography) supporting the
18
vineyard. Selectively harvesting blocks in response to this variability may be highly profitable. Whilst it
19
has also been shown that crop maturation is spatially variable, there may also be temporal variations
20
in the rate of maturation. Integrating knowledge of how vineyard spatial and temporal variation interact
21
has not previously been attempted and is the key objective in this work.
M.C.T Trought1 and R.G.V. Bramley2,3
1
New Zealand Institute for Plant & Food Research Ltd, Marlborough Wine Research Centre, PO Box
845, Blenheim, New Zealand; 2CSIRO Food Futures Flagship; 3CSIRO Sustainable Ecosystems, PMB
2, Glen Osmond, SA 5064, Australia;
1
Corresponding author: [email protected]
Abstract
Background and aims:
22
23
Methods and Results:
24
We used a proximal sensor to map vine vigour at high spatial resolution in a 5.9 ha Marlborough
25
vineyard planted to Sauvignon Blanc. Vigour measurements were also related to fruit soluble solids,
26
titratable acidity and pH - key indices of crop maturity. A knowledge of crop phenology and maturation
27
was used to predict how these indices changed with time. The pooled opinion of over 50 Marlborough
28
winemakers on the optimum levels of soluble solids, juice pH and titratable acidity at harvest to
29
produce a “typical Marlborough Sauvignon Blanc”, was used to develop a juice index which in turn
30
was mapped in space and time at the study site. The juice index showed marked spatial and temporal
31
variation.
32
33
Conclusions:
34
In addition to being spatially variable, grape quality in vineyards is also temporally variable. Thus, the
35
optimisation of decisions about harvest timing requires a knowledge of spatial variability. Conversely,
36
strategies such as selective harvesting cannot be properly optimized without a knowledge of crop
37
phenology – which is also spatially variable. In this study, we have shown that by integrating a
38
knowledge of crop phenology with an understanding of vineyard variability and winemaker objectives,
39
it is possible for the optimum harvest decision to be made such that fruit destined for a particular end-
40
use is harvested at the right time and from the right place.
41
42
Significance of the Study:
43
This is the first study in which knowledge of both spatial and temporal vineyard variation has been
44
integrated. It demonstrates that in order to be optimal, strategies such as selective harvesting need to
45
incorporate knowledge of crop phenology rather than relying on knowledge of spatial variation alone.
46
47
Keywords: Wine quality, phenology, terroir
48
49
50
Introduction
51
52
The importance that fruit composition has in determining the flavour and aroma characteristics of wine
53
was summarised by Rankin (1991) when he said that “the potential quality of wine is established in the
54
vineyard and carried to fruition in the winery”. Identifying and predicting the date when the balance of
55
flavour and aroma precursors are at their optimum could be considered the “holy grail” of viticultural
56
research. Conceptually, growers and winemakers will have harvest targets in respect of attributes
57
such as yield, disease status, other berry damage, soluble solids, acidity, colour and more recently
58
flavour and aroma characteristics. These targets may be modified, based on weather predictions and
59
harvesting infrastructure. Some targets may be well defined (in particular disease status, soluble
60
solids and yield) and may form the basis for payment schedules from wineries; other targets are less
61
clear. In addition, juice analyses are usually described as mean values, and fruit sampling protocols
62
attempt to achieve the best estimate of mean vineyard fruit composition. However, vineyards are
63
variable and identifying the target fruit composition becomes more complicated when the differences
64
in vine growth and fruit composition within a vineyard are considered. When the variation in
65
composition around the mean is large, it is likely that over- and under-ripe flavours are present in the
66
juice and subsequent wine (Trought 1997). The importance this has in determining overall juice quality
67
largely depends on the variety being considered. For example, typical Marlborough Sauvignon Blanc
68
displays “ripe” and “unripe” characteristics in the wine (Parr et al. 2007) However, the same unripe
69
character in a Cabernet Sauvignon may be considered to be inferior. Whatever the relationship
70
between variability in fruit composition and final wine quality, understanding and predicting the
71
variability is important if vineyards are to be managed appropriately and fruit harvested at a ripeness
72
to make the target wine style.
73
74
Recent research using precision agriculture techniques has provided tools to understand variability in
75
vineyards (Bramley and Hamilton, 2004, 2007; Proffitt et al., 2006; Tisseyre et al., 2007). To date this
76
knowledge has been used by viticulturists and/or winemakers to either differentially manage parts of
77
the vineyard to reflect this variability or to selectively harvest areas within the vineyard (Bramley, 2010;
78
Bramley and Hamilton 2005, 2007; Trought et al., 2008) and optimise the composition of parcels of
79
fruit (Bramley et al., 2005; Bramley and Hamilton, 2007). Hitherto, selective harvesting strategies have
80
relied on knowledge of spatial variation alone. Typically, they involve either harvesting a block on a
81
single occasion into two or more product streams and/or harvesting different parts of a block at
82
different times. In the latter case, the objective may be either to enable allocation of the fruit from the
83
entire block to the same product stream, or to maximise the match of the harvested parcels to their
84
intended end use (Bramley et al., 2005). By understanding how fruit composition is changing in
85
different areas of the vineyard, it may be possible for such harvesting decisions to be optimised in both
86
time and space.
87
88
This alternative approach requires an understanding of the differences in fruit composition throughout
89
the vineyard and how composition changes with time. Identifying the optimum date for harvest can
90
then be determined from the mean value and the variability around that mean. To do this, the
91
influence of various juice components on the overall value of the juice have to be defined, while at the
92
same time, changes in fruit composition with time need to be understood along with the impact on
93
these of vineyard factors such as soil texture or depth.
94
95
Trought (1997), Bramley and Hamilton (2007), Trought et al. (2008) and most recently Bramley et al.
96
(2010) have previously drawn attention to the variability of vineyards in the Marlborough region of New
97
Zealand, and the likely role that soil variation plays in driving this. In particular, the active nature of the
98
Wairau River flood plain has resulted in a pattern of silty hollows dissecting the sandier, more gravelly
99
soils which predominate (Rae and Tozer, 1990). As these hollows run predominantly in an
100
approximately east-west direction, whilst row orientation is generally north-south, the full range of
101
within-vineyard variation is commonly observed in a single row. This presents grape growers and
102
winemakers with a difficult challenge in regard to issues such as fruit parcelling and the scheduling of
103
harvest.
104
105
To our knowledge, the extent to which spatial variation in grape composition in a vineyard changes
106
with time has not been investigated before. Trought et al. (2008) measured trunk circumference and
107
pruning weight to characterise vine vigour and related variation in these attributes to variation in fruit
108
quality and soil properties (texture and temperature). In a companion paper (Bramley et al. 2010),
109
spatial analysis was used to identify relationships between trunk circumferences, soil texture and
110
canopy Plant Cell Density (PCD) measured using both remote and proximal sensing. In this paper we
111
explore how the spatial analysis of Bramley et al. (2010) could be integrated with models describing
112
aspects of phenological development and juice quality to better understand vineyard variation in
113
Marlborough, and thus, evaluate the likely commercial opportunities which might arise from such
114
understanding, particularly through improved harvest management.
115
116
117
Methods and Materials
118
119
The 5.9 ha trial site formed part of a commercial vineyard on the north side of the Wairau Plain
120
(173o35’E, 41o29’S). Details can be found elsewhere (Trought et al. 2008), but briefly, Sauvignon
121
blanc vines (clone UCD1 on SO4 rootstock) were planted in 1994 in rows, approximately orientated
122
north-south; the row spacing was 2.4m with a vine spacing in the row of 1.8m. The vines used in this
123
study were pruned using a 2-cane vertical shoot positioning system (Smart and Robinson, 1991),
124
retaining 12 nodes per shoot. Foliage wires were used to maintain a narrow canopy approximately
125
0.4m wide and 1.6m tall and vines were trimmed to maintain these dimensions three times during the
126
growing season. Vines were drip irrigated and pest and disease management was achieved by
127
following
128
circumference of all of the vines in four rows was measured 10cm above the graft union and 10cm
129
below the head of the vine. Six plots, (comprising of four vine bays) in each row were chosen to give a
130
range of circumferences ranging from 165 to 220cm.
131
NZ
Sustainable
Winegrowing
practice
(http://www.nzwine.com/swnz/).
The
trunk
132
Fruit composition was monitored approximately weekly from 18th February 2006 (just before veraison)
133
to harvest. Berry samples (32 berries) were collected from each plot, cooled in an insulated box and
134
analysed in the laboratory where fruit was gently macerated by hand, coarsely sieved and the juice
135
prepared for analysis. The soluble solids (oBrix) was determined using an Atago PAL-1 digital pocket
136
refractometer and pH was measured using a Metrohm 744 pH meter. The titratable acidity (TA) was
137
measured on a juice sample (2.5 mL until TA ~ 15 g/L after which time a 5.0 mL sample was used)
138
using a Metler Toledo DL50 autotitrator. The juice was diluted with 50 mL distilled water and titrated
139
using 0.1 M NaOH to pH 8.4. The titratable acidity is expressed as g/L tartatic acid equivalent.
140
141
A range of spatial data were collected and mapped to aid understanding of spatial variability within the
142
block (Bramley et al., 2010). These included the canopy plant cell density (PCD; the ratio of infrared:
143
red reflectance) measured using both airborne remote sensing, proximal sensing from the side using a
144
Crop Circle™ sensor (Holland Scientific, Lincoln, NE, USA), apparent electrical soil conductivity
145
(EM38; Geonics, Canada) and the measurements of trunk circumference described above. Of
146
particular importance to the present study are the Crop Circle data collected in March 2007 (Figure 1),
147
which were shown in our earlier study (Bramley et al., 2010) to closely mimic the other measures of
148
vine vigour (remote sensing, measurement of trunk circumference), irrespective of when these
149
measures were collected. Thus, Bramley et al. (2010) demonstrated that, in the same vineyard as
150
used for the present study, proximally sensed PCD provides a useful characterisation of vineyard
151
variability at high spatial resolution. Importantly, the spatial structure in a map of trunk circumference
152
derived from measurements made on the same target vines used here closely matched that in the
153
higher resolution PCD map. It should also be noted that, in spite of the changes in vine vigour shown
154
in Figure 1, yield differences over the block were small and not related to other measures of spatial
155
variability (Bramley et al., 2010).
156
157
158
Modelling changes in juice composition
159
160
Curves were fitted to juice analysis data from each plot using Sigma Plot v9 (Systat Software Inc.,
161
Chicago, IL, USA). A Gompertz 3-parameter curve was chosen to describe changes in soluble solids
162
as this provided a better model of soluble solids in a number of the plots when compared to a
163
quadratic polynomial. Quadratic polynomials were fitted to pH and titratable acidity. Interpretation of
164
these curves enabled us to estimate juice composition for each of the plots in the vineyard on any date
165
based on measurements made on 9 occasions during the season. These modelled values were used
166
in the subsequent analyses.
167
168
The PCD values for each plot were extracted from Figure 1 and related to the soluble solids, pH and
169
titratable acidity on a range of dates from shortly after veraison (Figure 3). A 3-parameter sigmoid
170
algorithm was fitted to the relationships between PCD and soluble solids and linear polynomials to the
171
relationships between PCD and titratable acidity and pH. Using this method we were able to use the
172
Crop Circle map (Figure 1) as a surrogate basis for the production of vineyard maps of these
173
components of grape juice and view how their values change on a daily bases. The daily values for
174
each of these components were also used to calculate the changes in the Juice Index Score (see
175
below) with time.
176
177
178
Development of a Juice index
179
180
Conceptually, growers and winemakers have juice composition targets for harvest. To develop a juice
181
index, 52 participants at a Sauvignon Blanc Seminar in Marlborough, held in November 2005, were
182
asked how they would rate juices to make a “typical Marlborough Sauvignon Blanc”. The participants
183
consisted of 17 grapegrowers or viticulturists, 31 winemakers and the balance were either consultants,
184
laboratory staff, or did not indicate their role, although overall, the 52 participants had an average
185
industry experience of 11.3 years.
186
187
Scales of soluble solids values (between 18 and 27 °Brix in 0.25° steps), titratable acidity (between 5
188
and 14 g/L in 0.25 g/L steps) and pH (between pH 2 and 5 in pH 0.1 steps) were presented to the
189
participants. These ranges covered values well to either side of the expected optima. Participants
190
were asked to give a value to each point on the scale where 1 was not preferred and 5 strongly
191
preferred. A Gaussian 4-parameter curve was fitted to each data set and from this, a preference score
192
(ps) could be determined for each value of soluble solids (SS), pH and titratable acidity (TA). The
193
participants were also asked to ascribe a weighting (w) to each of the three components when making
194
their harvest decision. These were used to develop the overall Juice Index (JI):
195
196
(1)
197
198
We recognise that other factors influence that decision, such as winery logistics and weather events
199
and a number of the respondents also introduced comments on flavour and fruit health. However, as
200
we were unable to quantify the intensity of these comments, they were not included in the analysis.
201
202
Using the modelled juice composition (as above) for each trial plot on each day during the later stages
203
of fruit ripening (from 22 March) and the relationships between these modelled values and Crop Circle
204
PCD (Figure 1), values of SSps, TAps and pHps and thus, JI were calculated for every pixel (2m x 2m) in
205
a map grid imposed over the vineyard area. These values were also used to determine the distribution
206
SS, pH and TA composition in the vineyard as a whole.
207
208
209
Results
210
211
Fruit collected from the 24 plots increased in soluble solids from 5.2 to 22.3 oBrix during the course of
212
the monitoring period (Figure 2). At the same time, the pH increased from 2.56 to 3.05 and TA
213
decreased from 38.9 to 9.3 g L-1 (Figure 2). However, the rate of change was not the same across the
214
whole vineyard. Fitting the Gompertz 3-parameter curves to soluble solids data and quadratic
215
polynomials to titratable acidity and pH values from each plot enabled us to estimate the fruit
216
composition for each plot on a daily basis. The daily values were used to estimate the variability and
217
map the vineyard.
218
219
Vines with low PCD (i.e. low vigour) had higher soluble solids and pH and lower TA early in the
220
ripening period (March 22), when compared with those of higher PCD (i.e. higher vigour; Figure 3). As
221
fruit ripened, the soluble solids became less variable, with all vines (except those of PCD greater than
222
10.7) achieving a soluble solids of approximately 23.0 oBrix. The predictive power of the PCD also
223
changed with time, reflecting the changes in the ripening profile (Figure 4). The PCD gave a good
224
prediction of differences in fruit soluble solids (an R2 of up to 0.75) early in ripening, but its predictive
225
utility then decreased as the slope of the relationship decreased (Figures 3, 4). This reflected the
226
slower rate of soluble solids accumulation by fruit where vines had low PCD (those growing on stony
227
soils) when compared to those with high PCD on soils with higher silt content. Similar changes were
228
observed in the relationships between PCD and pH and TA. Vines with low PCD generally had lower
229
TA and higher pH early in the ripening period, but differences between the plots decreased to low
230
levels later in ripening (Figure 3). The predictive ability of PCD for TA and pH was less robust than
231
soluble solids, but interestingly, increased later in the ripening period (Figure 4). The changes in the
232
relationships between PCD and TA and pH values across the vineyard (Figure 4) indicate that at times
233
during ripening, vine to vine differences in these fruit components were unrelated to vine vigour as
234
measured by PCD and/or trunk circumference. We recognise that the fit of our regressions has low
235
predictive ability on some dates, but persist with this approach here as a means of developing the
236
proof of concept. The important factor to note is that the ability to predict fruit composition is best early
237
in the ripening of the fruit, which might be valuable for the early planning of harvest strategies or
238
product allocation.
239
240
The differences in fruit composition were reflected in the spatial variability of the vineyard (Figures 5-
241
7), with vines on the eastern side of the vineyard, generally achieving a higher soluble solids and pH,
242
and lower TA earlier than those on the west. In general, the variability in soluble solids decreased
243
across the whole vineyard as fruit ripened (Figure 8). The distribution was skewed to the left
244
throughout the ripening period, although a maximum value of between 23.5 and 24 oBrix was achieved
245
by 1st May (Figure 8). The pH and TA achieved a minimum degree of variability on April 21, and then
246
appeared to become more variable as fruit ripened further (Figure 8).
247
248
The optimum values for the juice components determined from the questions posed to the Sauvignon
249
Blanc workshop participants were 22.5 oBrix, TA 9.0 g/L and pH 3.2 (Figure 9) with the weighting given
250
to each component in considering harvest decisions being 49.3, 27.7 and 23.0 % respectively. These
251
data enabled mapping of changes in juice index (equation 1) as the crop matured (Figure 10). As a
252
consequence of the earlier ripening profile of fruit on the (north) eastern side of the vineyard, this area
253
reached its maximum juice score earlier than elsewhere. In this part of the block, the juice index
254
generally reached a value of >4.25 between the 11 and 21 April (Figure 10). However, by 1 May, the
255
soluble solids was greater than 22.5 oBrix and as a result the juice index started to decrease. In
256
contrast, the slower ripening on the western side of the vineyard, meant that the juice index failed to
257
reach 4.25 by 1 May. The influence that the increasing soluble solids has on juice index distribution is
258
demonstrated in Figure 11, which shows that by May 1, the mean index value had decreased from an
259
average of 4.2 back to 3.3 as soluble solids became too high.
260
261
262
Discussion
263
264
Despite the potential importance that differences in maturity class within a harvest sample potentially
265
have for a final wine, it has seldom been demonstrated. Carroll et al (1978) sorted Carlos Muscadine
266
grapes by colour using a photoelectric light source and concluded that wines made from optimum
267
ripeness were superior to those made from under- or over-ripe fruit. However, the importance of
268
variation in fruit composition on wine style may depend on the variety being considered. For example,
269
Long (1987), compared wine quality from two Californian vineyards and reported that the vineyard with
270
the least variation in ripeness consistently produced the better quality wine. In contrast, Parr et al
271
(2007) suggested that “archetypical” Marlborough Sauvignon Blanc has both “ripe” and “unripe”
272
flavours in the wine. This suggests that regardless of the wine style, it is important to understand and
273
be able to predict both the variability in fruit composition in a harvest sample, and also the range of
274
wine aromas and flavours that are likely to be derived from the fruit.
275
276
Identifying the optimum time to harvest grapes in a variable vineyard is a challenge facing most
277
viticulturists and winemakers. Whether a vineyard is harvested or not is a decision made by
278
viticulturists and winemakers, often based on the question: “do I think this fruit will be better next time
279
the logistics of winery intake will allow the fruit to be processed?”. Viticulturists and winemakers
280
generally have compositional targets, which in some cases may form the basis of winery payments for
281
grapes. To integrate various common juice parameters, we developed the concept of a “juice index”.
282
This enables fruit with a wide range of composition to be compared against a single scale.
283
Traditionally, fruit is sampled to obtain a representative value of key maturity indices (e.g. soluble
284
solids, pH and TA) and little regard is given to either the variability around that mean, or the location of
285
sample collection; representativeness is assumed. However, as vineyards are variable some areas
286
may be unripe while others are overripe. Wines made from these juices will have different flavour and
287
aroma profiles. For example, as soluble solids levels increase, the resulting wines may have higher
288
alcohol concentrations. In contrast, unripe fruit may have excessively high acidity and/or herbaceous
289
character.
290
291
By measuring changes in composition with time, we have modelled how soluble solids, TA and pH
292
change in a variable vineyard in an attempt to identify the optimum harvest date. Our results show that
293
early in ripening, considerable variation in fruit composition is observed and this is related to other
294
measures of variability (e.g. PCD and trunk circumference). Vines with lower PCD had higher soluble
295
solids and pH and lower titratable acidity and ripened earlier when compared to vines with higher PCD
296
values. The vines with low PCD had smaller trunk circumferences and were growing in soils with
297
greater proportions of stones in the profile (Bramley et al. 2010; Trought et al. 2008; Mills 2006). In
298
contrast, vines growing on deeper, siltier soils in the lower lying parts of the vineyard failed to achieve
299
the target maturity values by May 1 (Figures 10, 12). Other areas did eventually ripen to target levels,
300
but the fact that the majority of the vines were now past their optimum ripeness level meant that the
301
overall value of the juice index decreased between April 21 and May 1. This indicates that, if the
302
vineyard was to be harvested as a single parcel, the optimum date to harvest fruit was about April 21
303
(in this particular season), when the overall variation in vineyard juice index was small and the mean
304
value was highest.
305
306
The modelled data clearly show the date and location of fruit with the highest juice index. For example
307
on April 21st 88.5% of the vineyard area had achieved a juice index of more than 4.4, with the balance
308
of the vineyard still unripe (Figure 12). However, by delaying harvest beyond this date, soluble solids
309
increased above the optimum concentration and juice index values decreased (Figure 10).
310
Understanding the spatial variability in a vineyard helps viticulturists make informed harvesting
311
decisions. Indeed, previous work has shown that selective harvesting, based on such knowledge, can
312
increase overall vineyard profitability (Bramley et al., 2005). In the case of our study block, such a
313
strategy might involve partitioning the block into two areas. The first area, which achieves a juice index
314
of 4.4 or greater (Figure 12), may be processed to a wine with a higher price point compared to the
315
remaining second area. However, as Figures 5-7 and 10-12 illustrate, this variability potentially
316
changes with time. This reflects differences in the rate of change in key volatile and non-volatile grape
317
components in various parts of the vineyard, reflecting differences in vine vigour, which in turn are a
318
consequence of factors such as soil texture (Bramley et al., 2010). By taking this spatial and temporal
319
variation into account, the optimum harvest date for the whole vineyard can be determined (Figure
320
12). Likewise, the consequence of delaying harvest to all or part of the vineyard on potential juice
321
quality can be evaluated. In our example, the fruit on the western side of the vineyard fails to reach a
322
juice index of 4.4 by early May, when short days and cool temperatures (and the risk of frost) means
323
that it is unlikely that fruit will ripen further.
324
325
Spatial variability in vine growth (e.g. trunk circumference) in this vineyard reflects variation in soil
326
texture (Mills, 2006; Trought et al., 2008; Bramley et al 2010) and is therefore unlikely to change
327
between seasons. However, the magnitude of variability in juice composition does change as the fruit
328
ripens. It is also likely that the relationships between PCD and fruit composition parameters will be
329
different between seasons and as a result the optimum date of harvest will change. By understanding
330
the spatial variability in the vineyard, the relative balance of soluble solids, pH and TA early in ripening
331
(shortly after veraison) and how season affects fruit composition, the modelling approach we used
332
may be able to predict the optimum harvest date. This will be investigated in a further paper.
333
334
It might be suggested that more robust estimates of the juice index and its spatial and temporal
335
variability may have been obtained had we measured indices of juice composition on a spatially
336
distributed array of target vines in the manner of Bramley (2005) rather than relying on the PCD map
337
(Figure 1) as the basis for extrapolating juice compositional data from our fewer, more spatially
338
restricted target vines. Such a strategy would certainly have enabled maps similar to those shown in
339
Figure 5-7 to be produced, but these would have had a greatly reduced ‘support’ (Webster and Oliver,
340
2007) compared to the PCD maps derived from (very) high resolution data. Indeed, the 24 target vines
341
used here are insufficient for geostatistical analysis and are obviously far fewer than the 41,660 PCD
342
values which underpin Figure 1. Furthermore, a practitioner is never going to have the time to monitor
343
more than a few target vines – certainly too few to produce maps from – yet they may buy some
344
remotely sensed imagery or rent (or even buy) a proximal sensor such as the one used here in order
345
to get the high resolution data which Figure 1 derives from; the collection of these data and
346
subsequent map interpolation requires just a few hours work.
347
348
Our results also have implications for the concept of terroir. Bramley and Hamilton (2007) have
349
previously raised questions as to the scale at which the concept of terroir is useful. They suggested
350
that there is a mismatch between the common regional-scale use of this concept and the within-
351
vineyard scale at which it might contribute to an understanding of the style and sensory attributes of a
352
wine as a consequence of the biophysical environment in which its source grapes were grown. Whilst
353
the present results are consistent with those of Bramley and Hamilton (2007), they also suggest a
354
temporal component to terroir which has arguably been ignored in the past. Thus, a wine reflects both
355
the location in which the source fruit were grown, and also the time at which that fruit was harvested.
356
In this connection, both ‘location’ and ‘time’ can be thought of as means about which there is
357
considerable variation (Figures 8 and 12).
358
359
360
Conclusions
361
362
For viticulturists and winemakers to anticipate the optimum harvest date for a vineyard, or to adopt
363
differential management strategies within it such as selective harvesting, a knowledge of both spatial
364
variability and its interaction with crop maturation is needed. In this study, we have shown that by
365
integrating our knowledge of crop phenology with an understanding of vineyard variability and
366
winemaker objectives, decisions about where and when to harvest can be optimised. Further work is
367
required to take this beyond the proof of concept, particularly with respect to the extrapolation of our
368
method to other locations, other varieties and growing seasons. However, whilst the objective here
369
was not to remove the sensory evaluation of the winemakers from the harvesting decision process,
370
our results suggest that an appropriate understanding of vineyard performance in both space and time
371
will help their decision processes at a busy time of the year.
372
373
374
Acknowledgments
375
376
This work was variously funded by the Foundation for Scientific Research (CO6X0707 – Designer
377
grapevines) and the Marlborough Wine Research Centre (IN 0502 – Influence of soil variation in a
378
Marlborough Sauvignon Blanc vineyard). The input of RGVB was made possible by the awarding of a
379
‘Liquorland Top 100 Fellowship’ through the Marlborough Wine Research Centre with additional
380
support from the CSIRO Food Futures Flagship. We are most grateful to Andy Frost, Andrew Naylor,
381
Rod Brailsford and colleagues (Pernod Ricard, New Zealand Ltd.) for their support of the work through
382
discussion, provision of vineyard access and assistance with yield mapping. In particular we also
383
thank Robyn Dixon (formerly MSc student at Lincoln University) for access to data from her Masters
384
thesis. We also acknowledge the useful discussion of Alistair Hall (Plant & Food Research,
385
Palmerston North) in fitting the curves and the helpful comments which Drs Richard Hamilton (Foster's
386
Wines), Bruce Searle (Plant and Food Research) and Paul Boss (CSIRO Plant Industry) made on an
387
earlier draft of this paper.
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
References
Bramley, R.G.V. (2005) Understanding variability in winegrape production systems. 2. Within vineyard
variation in quality over several vintages. Australian Journal of Grape and Wine Research 11, 33-42.
Bramley, R.G.V. (2010) Precision Viticulture: Managing vineyard variability for improved quality
outcomes. Chapter 13 in: Reynolds, A.G. (Ed) Understanding and Managing Wine Quality and
Safety. (Woodhead Publishing: Cambridge, UK). In press.
Bramley, R.G.V. and Hamilton, R.P. (2004) Understanding variability in winegrape production systems.
1. Within vineyard variation in yield over several vintages. Australian Journal of Grape and Wine
Research 10, 32-45.
Bramley, R.G.V. and Hamilton, R.P. (2005) Hitting the zone – Making viticulture more precise. In: Blair,
R.J, Williams, P.J. and Pretorius, I.S. (Eds) Proceedings of the 12th Australian Wine Industry
Technical Conference. Winetitles, Adelaide. pp. 57-61.
Bramley, R.G.V. and Hamilton, R.P. (2007) Terroir and Precision Viticulture: Are they compatible?
Journal international des Sciences de la Vigne et du Vin 41, 1-8.
Bramley, R.G.V., Trought, M.C.T. and Praat, J-P. (2010). Vineyard variability in Marlborough, New
Zealand: Characterising variation in vineyard performance and options for the implementation of
Precision Viticulture. Australian Journal of Grape and Wine Research. Submitted.
Bramley, R.G.V., Proffitt, A.P.B., Hinze, C.J., Pearse, B. and Hamilton, R.P. (2005) Generating benefits
from Precision Viticulture through selective harvesting. In: Stafford, J.V. (Ed) Proceedings of the 5th
European Conference on Precision Agriculture. Wageningen Academic Publishers, The
Netherlands. 891-898.
Carroll, D.E., Ballinger, W.E., McClure, W.F. and Nesbitt, W.B.. (1978). Wine quality versus ripeness of
light-sorted Carlos Muscadine grapes. American Journal of Enology and Viticulture 29, 169-171.
Long, Z.R. (1987). Manipulation of grape flavour in the vineyard: California, North Coast region. In:
Proceeding of the Sith Australian Wine Industry Conference, Adelaide. T.H. Lee (Ed.) pp. 82-88
Mills, T.S. (2006) Relations among geology, soil type and Sauvignon Blanc vineyard variation in
Marlborough, New Zealand. MSc Thesis. University of Auckland.
Parr W.V., Green, J.A., White, K.G. and Sherlock, R.R. (2007). The distinctive flavour of New Zealand
Sauvignon blanc: Sensory characterisation by wine professionals. Food Quality and Preference, 18,
849-861.
Proffitt, T., Bramley, R., Lamb, D. and Winter, E. (2006) Precision Viticulture – A new era in vineyard
management and wine production. (Winetitles: Adelaide).
Rae, S. N. and Tozer, C. G. (1990) Land and Soil Resources. Blenheim, Nelson-Marlborough Regional
Council.
Rankin, B. (1991). Madking good wine. PanMacmillan publishers, Australia
Smart, R. E. and Robinson, M. D. (1991) Sunlight into wine. Winetitles, Adelaide.
Tisseyre, B., Ojeda, H. and Taylor, J. (2007) New technologies and methodologies for site-specific
viticulture. Journal international des Sciences de la Vigne et du Vin 41, 63-76.
Trought, M.C.T. (1997) The New Zealand terroir: sources of variation in fruit composition in New Zealand
vineyards. In: Henick-Kling, T., Wolf, T.E. and Harkness, E.M. (Eds) Proceedings of the 4th
international symposium on cool climate viticulture and enology, Rochester, New York, USA, 16-20
July 1996. I 23-27.
Trought, M.C.T., Dixon, R., Mills, T., Greven, M., Agnew, R., Mauk, J.L. and Praat, J-P. (2008) The
impact of differences in soil texture within a vineyard on vine vigour, vine earliness and juice
composition. Journal international des Sciences de la Vigne et du Vin 42, 67-72.
Webster, R. and Oliver, M. A. (2007). Geostatistics for Environmental Scientists. Wiley and Sons,
Chichester.
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
List of Figures and Tables
Figure 1.
Canopy plant cell density (PCD; the ratio of infrared:red reflectance) as measured
using a Crop Circle sensor in March 2007. (Data of Bramley et al., 2010).
Figure 2.
Changes in soluble solids, titratable acidity and pH of Sauvignon Blanc from the 24
plots used in the experiment. Veraison date was assessed to be when 50% of the
berries had softened (8.5 oBrix). Error bars are +/- one standard error.
Figure 3.
Relationships between Plant Cell Density and soluble solids, titratable acidity and pH
of Sauvignon Blanc on different dates during fruit ripening.
Figure 4.
Changes in correlation coefficient PCD v soluble solids, titratable acidity and pH at
different times during fruit ripening.
Figure 5.
Spatial and temporal differences in grape soluble solids.
Figure 6.
Spatial and temporal differences in grape titratable acidity.
Figure 7.
Spatial and temporal differences in grape juice pH.
Figure 8.
Changes in the soluble solids, titratable acidity and pH distribution in the Sauvignon
Blanc vineyard with time
Figure 9.
Soluble solids (a), titratable acidity (b) and pH (c) value scores. Fifty two experienced
industry personnel were asked how they would rate juice with soluble solids ranging
from 18 to 27 oBrix, titratable acidity from 5 to 14 g/L and pH from 2.0 to 4.5 to make a
“typical Marlborough Sauvignon Blanc”. 1= poor; 5= good.
Figure 10.
Spatial and temporal changes in juice index with time in a Sauvignon Blanc vineyard.
Figure 11.
Modelled variation in juice index in a Sauvignon Blanc vineyard on different dates.
Figure 12.
Maximum juice index achieved and the date on which a score of 4.4 or better was
reached. The % values listed in the legend to the right hand map represented the
proportion of the vineyard area falling into that category.
479
480
481
482
Figure 1.
Canopy plant cell density (PCD; the ratio of infrared:red reflectance) as measured
using a Crop Circle sensor in March 2007. (Data of Bramley et al., 2010).
3.2
25
30
15
Soluble solids
Titratable acidity
pH
10
20
5
10
Veraison
0
3.0
2.9
2.8
2.7
2.6
2.5
20
483
484
485
486
487
488
489
3.1
pH
20
Titratable acidity (g/L)
Soluble soids (oBrix)
40
40
60
80
Days from 1 February 2006
Figure 2.
Changes in soluble solids, titratable acidity and pH of Sauvignon Blanc from the 24
plots used in the experiment. Veraison date was assessed to be when 50% of the
berries had softened (8.5 oBrix). Error bars are +/- one standard error.
18
22 March
24
3.2
16
22
14
3.1
20
12
3.0
10
18
2.9
8
16
6
24
4
18
29 March
2.8
2.7
3.2
16
Titratable acidity (g/L tartaric acid)
Soluble solids (oBrix)
20
18
16
5 April
24
22
20
18
16
12 April
24
22
14
3.1
12
3.0
10
2.9
8
2.8
6
2.7
4
18
pH
22
16
3.2
14
3.1
12
3.0
10
2.9
8
6
2.8
4
18
2.7
16
3.2
14
3.1
12
20
2D Graph 36
3.0
10
18
2.9
8
16
6
2.8
2.7
24
4
18
16
3.2
14
3.1
22
12
21 April
20
3.0
10
Crop circle v Soluble solids (1st May)
18
2D Graph 7
2.9
8
2.8
6
16
4
18
24
22
20
1 May
2.7
16
3.2
14
3.1
12
3.0
10
18
8
16
6
2.9
2.8
4
8.5
490
491
492
493
494
9.0
9.5
10.0
10.5
11.0
11.5
2.7
8.5
9.0
9.5
10.0
10.5
11.0
11.5
8.5
9.0
9.5
10.0
10.5
11.0
11.5
Crop circle PCD
Figure 3.
Relationships between Plant Cell Density and soluble solids, titratable acidity and pH
of Sauvignon Blanc on different dates during fruit ripening.
R2 PCD v juice component
0.8
0.6
0.4
0.2
0.0
22 Mar
495
496
497
498
499
Soluble solids (oBrix)
Titratable acidity (g/L)
pH
1 April
11 April
21 April
1 May
Date
Figure 4.
Changes in correlation coefficient PCD v soluble solids, titratable acidity and pH at
different times during fruit ripening.
500
501
502
503
Figure 5.
Spatial and temporal differences in grape soluble solids.
504
505
506
Figure 6.
Spatial and temporal differences in grape titratable acidity.
507
508
509
510
Figure 7.
Spatial and temporal differences in grape juice pH.
100
80
2nd March
2nd March
2nd March
60
40
20
0
80
60
12th March
12th March
40
20
0
Proportion of vineyard in each category
80
60
22nd March
22nd March
40
20
0
80
60
1st April
1st April
40
20
0
80
11th April
60
40
11th April
20
0
80
21st April
60
40
21st April
20
0
80
1st May
60
40
1st May
20
0
5
511
512
513
514
515
516
10
15
20
Soluble solids (oBrix)
Figure 8.
25 5
10
15
20
25
Titratable acidity (g/L)
30 2.4
2.6
2.8
3.0
3.2
3.4
pH
Changes in soluble solids, Titratable acidity and pH distribution in the Sauvignon
Blanc vineyard with time
517
5
a
4
3
2
1
18
20
5
24
26
b
4
Value score
22
Soluble solids (oBrix)
3
2
1
6
8
10
12
14
Titratable acidity
5
c
4
3
2
1
2.0
518
519
520
521
522
Figure 9.
2.5
3.0
3.5
4.0
4.5
5.0
Juice pH
Soluble solids (a), titratable acidity (b) and pH (c) value scores. Fifty two experienced
industry personnel were asked how they would rate juice with soluble solids ranging
from 18 to 27 oBrix, titratable acidity from 5 to 14 g/L and pH from 2.0 to 4.5 to make a
“typical Marlborough Sauvignon Blanc”. 1= poor; 5= good.
523
524
525
526
Figure 10.
Spatial and temporal changes in juice index with time in a Sauvignon Blanc vineyard.
120
12 March
22 March
1 April
11 April
21 April
1 May
Proportion of vineyard
100
80
60
40
20
0
1
527
528
529
530
2
3
4
5
Juice Value Index
Figure 11.
Modelled variation in juice index in a Sauvignon Blanc vineyard on different dates.
531
532
533
534
535
Figure 12.
Maximum juice index achieved, the date on which a score of 4.4 or better was reached
and the relationship of these with vine vigour (PCD) and elevation.