Content uploaded by Wies U Cynkar
Author content
All content in this area was uploaded by Wies U Cynkar on Feb 09, 2014
Content may be subject to copyright.
D. Cozzolino et al., J. Near Infrared Spectrosc. 14, 279–289 (2006) 279
© NIR Publications 2006, ISSN 0967-0335
Introduction
Wine is one of the most ancient beverages produced by
man. It is made by transforming sugars into alcohol dur-
ing fermentation of grape must. Wine is composed mainly
of water (70–90%), ethanol (8–20%), sugars (0.1–20%)
and acids.
1
However, the wine matrix also contains other
chemical compounds that vary in concentration (i.e. from %
to ppm) that might greatly infl uence the sensory properties
of the fi nal product.
1
To enhance the quality of its products further, the wine
industry has a clear need for simple, rapid and cheap tech-
niques for objectively evaluating the quality of grapes. The
measurement of grape characteristics which impact on
product quality is a basic requirement for vineyard improve-
ment and for optimum production of desired wine styles. It
is common industry practice for quality assessment to be
achieved by total soluble solids (TSS) and acidity measure-
ment, visual assessment and also by tasting assessment of
fruit and of wines following vinifi cation. Acidity and solu-
ble solid measurements are insuffi cient as quality indicators
and it is not possible to adequately assess quality by tasting
alone. However, there is a strong need in the modern wine
industry for timely information that can be used for grape
berry maturity assessment, identifi cation of vineyard blocks or
sections of a vineyard that should be segregated and load qual-
ity assessment. Existing analytical methods for measurement
of grape composition are not appropriate for the demands of
production; the world grape and wine production in 2001 was
estimated to be around 61,225 million tonnes and 26,473 ML,
respectively.
2
Therefore, the two factors of rapidity and low
cost of analysis are of paramount importance in the modern
wine industry. Even simple analyses currently require sam-
ples to be sent to a geographically separate laboratory, with
inherent delays in achieving results. More complex analyses,
such as grape colour measurement, terpene analysis, nitrogen
concentration and phenolics analysis are not considered as
serious options by the industry because of their cost and slow
turn around time.
3
Assessment of juice, wine and spirit quality
is also restricted at present for similar reasons.
Spectroscopic techniques such as near infrared (NIR)
spectroscopy offer possibilities for simple, rapid and cost-
effective analysis throughout the wine industry production
chain, starting with grapes and fi nishing with wines and
Review
Analysis of grapes and wine by near infrared
spectroscopy
D. Cozzolino,
*
R.G. Dambergs, L. Janik, W.U. Cynkar and M. Gishen
The Australian Wine Research Institute, Waite Road, Urrbrae, PO Box 197, Adelaide, SA 5064 and The Cooperative Research Centre
for Viticulture, PO Box 154,Glen Osmond, SA 5064, Australia
Spectroscopic techniques such as near infrared (NIR) spectroscopy are used in the food industry to monitor and assess the
composition and quality of products. Similar to other food industries, the wine industry has a clear need for simple, rapid and
cost-effective techniques for objectively evaluating the quality of grapes, wines and spirits. Thirty years have passed since the
fi rst work reported by Kaffka and Norris on the use of NIR spectroscopy to analyse wine. Since then, NIR spectroscopy has
been used for grape and wine compositional analysis, fermentation monitoring and wine grading. However, the use of NIR
spectroscopy in the wine industry is still in its infancy. From the analysis of the scatter information available, it appears that NIR
spectroscopy is applied in different steps during the wine production. This review highlights the most recent applications of NIR
spectroscopy in the grape and wine industry. Additional information is also provided on the use of mid infrared spectroscopy
for wine analysis.
Keywords: wine, grapes, composition, near infrared, visible, chemometrics, fermentation
280 Review: Analysis of Grapes and Wine by NIR
sprits.
3
The use of NIR spectroscopy in the wine industry
dates back to some early work by Kaffka and Norris.
4
Their
preliminary work was performed on a relatively small
number of test samples (n = 26) prepared by standard addi-
tion of some of the main components of interest (for example,
ethanol, fructose and tartaric acid) to a red and a white wine.
These samples were scanned in transmission mode using
three path lengths (0.3, 1 and 5 mm path length). Although
these samples represented alterations within the same two
basic wine matrices, they allowed the identifi cation of criti-
cal wavelengths that could be utilised for multiple linear
regression (MLR) analysis.
Nowadays, the main application of NIR spectroscopy
in the wine industry is the measurement of alcohol content
using fi lter instruments with two or three wavelengths.
4–8
However, since the availability of new instruments (for exam-
ple, monochromator or diode array spectrophotometers), the
availability of faster computers and the development of new
algorithms and software for chemometric analysis, all the
information residing in a full spectrum and new applica-
tions of NIR spectroscopy in the wine industry are now
evidenced.
9
This review highlights the most recent applications of
NIR spectroscopy to analyse the composition of grapes and
wine. Additional information is also provided on the use of
mid infrared (MIR) spectroscopy for wine analysis.
Measurement of grape composition
Grape composition at harvest is one of the most
important factors that determine the future quality of the
wine.
10–14
Traditionally, grapes are harvested based on the
concentration of TSS, which is a measure of sugars, mainly
glucose and fructose, determined using a refractometer.
12,13
The prediction of quality parameters in red grapes using
NIR spectroscopy is usually conducted by scanning
homogenised grape samples using a research grade
laboratory NIR spectrophotometer,
3,10,11
but other sample
presentation modes (including whole grapes) and cheaper,
more adaptable instruments have been used.
15–25
It is well
known that NIR spectroscopy is able to measure TSS in
other fruits
7
and several authors have reported the use of
NIR spectroscopy to measure TSS in grapes and must.
3,11–
17,25,26
Table 1 shows the standard error of prediction (SEP)
obtained for several chemical compositional parameters
in grapes reported by various research groups. A wide
range of SEP values have been obtained, depending on
the variety and the wavelength region used to develop
the calibration models. Ranges of wavelength used to
measure TSS in grapes were from 900 to 1050 nm,
18
300
to 1160 nm,
25
700 to 2500 nm,
15
400 to 2500
11
and 650 to
1100 nm.
16
Due to the fact that most of the studies used PLS
as the regression method and the lack of available data
provided by the different authors, it was diffi cult to defi ne
a specifi c wavelength best suited for predicting TSS in
Reference Parameter Number of samples Varieties Wavelength range (nm) Mode Range
SEP RPD
Jaren et al. (2001)
15
TSS 30 GR and V 800–2500 R 14–28 1.04–1.59 (°Brix) n/a
Herrera et al. (2003)
16
TSS > 200 CS, C, CH 650–1100 R, T 7–30 1.34–2.96 (°Brix) 3.0–4.0
Dambergs et al. (2003)
10
Total anthocyanins > 3000 CS, SH, M, GR 400–2500 R 0.2–2.6 0.05 (mg g
–1
) 3.8
pH R 2.8–4.9 0.11 2.8
TSS R 10.7–37.4 1 (°Brix) 10.3
Chauchard et al. (2004)
26
Acidity
*
371 CG, MV and UB 680–1100 R 2–16 1.28(g L
–1
) n/a
Arana et al. (2005)
25
TSS V and CH 500–800 R 15–26 1.27 (°Brix) 1.33–1.88
Cozzolino et al. 2004
23
Total anthocyanins 60 CS, SH and M 400–1100 R 0.4–1.7 0.06 (mg g
–1
) 4.2
pH R 3.6–4.3 0.045 2.2
*
acidity (malic + tartaric acid), R: refl ectance, T: transmittance,SEP: standard error of prediction, TSS; total soluble solids, RPD: SD/SEP, n/a: no available data to calculate RPD,
CG: Carignan, C: Carmenere, M: Merlot, CH: Chardonnay, CS: Cabernet Sauvignon, SH: Shiraz, GR: Grenache, V: Viura, UB: UgniBlanc, MV: Mourverdre.
Table 1. Analysis of grape composition by visible and near infrared refl ectance spectroscopy.
D. Cozzolino et al., J. Near Infrared Spectrosc. 14, 279–289 (2006) 281
grapes. However, it seems that the calibrations developed
for TSS use wavelengths that are related to O–H and C–H
bonds, around 980 nm, 1400 nm, 1900 nm and 2170 nm.
7
The use of in-line sensors to assess sugars and moisture
content in red grapes has also been reported and an early
study showed that a standard error of calibration (SEC)
range of 0.5 to 3.9 °Brix for the in-line prediction of sugars
in grape must can be achieved.
18
It has been reported that grape total anthocyanin
concentration (colour) is a good predictor of red wine
composition and quality and is widely used by the
Australian wine industry.
10–13,19–24
Several techniques have
been evaluated by the wine industry to objectively assess
the composition of grapes and wines for quality control and
payment purposes.
19–24
It has been demonstrated that total
anthocyanins, TSS and pH can be measured with partial
least squares (PLS) regression using refl ectance spectra of
homogenates of red grape berries scanned over the wave-
length range of 400–2500 nm.
3,10
With calibrations for total
anthocyanins in red grapes, it has been observed that for
large data sets incorporating many vintages, regions and
grape varieties, PLS calibrations show pronounced non-
linearity.
10,20–24
The SEP values reported for total anthocy-
anins vary from 0.05 to 0.18 mg g
–1
and increase with diverse
sample sets in comparison to sample sets restricted on the
basis of growing region and/or variety. These observations
may be related to non-linearities observed in the calibra-
tions produced with the diverse sets.
10,20–24
The prediction
accuracy using calibrations derived from restricted sample
sets approached that of the reference methods for total
anthocyanins and pH. An alternative strategy to mitigate
the effects of non-linearity on the NIR calibrations for total
anthocyanins is to use LOCAL regression.
10,17,20,22,27
The
same studies demonstrated that TSS calibrations were not
signifi cantly affected by the sample matrix.
10,17,20,22
The use of
NIR spectroscopy has now been put into practice by several
large Australian wine companies for the determination
of the concentration of total anthocyanins (colour) in red
grapes for payment purposes. In 2003, it was estimated that
20% of the total crush in Australia used visible (vis) and
NIR spectroscopy to assess the composition of the grapes
harvested.
10
Similar NIR applications have been reported
by private wineries and research groups in Chile, Spain and
Portugal.
15,16,25
The possibility of simplifying the sample presentation (for
example, using whole grapes instead of homogenates) could
dramatically increase sample throughput in the winery.
23
Scanning of single berries is also a possibility; however,
high coeffi cients of variation (40%) in the visible (vis)-NIR
spectra were observed when samples were rotated or scanned
in different positions relative to the spectrophotometer.
9,16,28
This variation within the same berry might be due to variations
in chemical composition, dust and to different degrees of sun
exposure (shading).
9,16,28
Investigations of whole grape berry
presentation using a diode array spectrophotometer indicated
that NIR may have potential for use at the weighbridge or
for in-fi eld analysis of total anthocyanins, TSS and pH.
23
The major challenges on the measurement of whole grapes
are related to sample presentation, instrument availability
and cost and with the desirable accuracy for the prediction
of chemical composition that might be achieved. Figure 1
shows the vis and NIR mean spectra of a single berry, whole
grape bunch and homogenates of grapes. Differences in the
spectra were observed around the O–H absorption bands
(980 nm, 1400 nm and 1900 nm) as well as in the vis region
(pigments).
7
The use of a diode array instrument to predict total acidity
(measured by HPLC as malic and tartaric acid) in a set of
white grape varieties has also been reported by researchers
in France.
26
In this study, several algorithms were used to
overcome non-linearity problems in the models developed.
The least squares vector machine (LS-SVM) algorithm
performed best. Comparing PLS and LS-SVM, the standard
error of cross-validation (SECV) and SEP were 1.32 and
1.28 g L
–1
for PLS and 1.11 and 1.03 g L
–1
for LS-SVM,
respectively.
26
Fungal diseases in grapes
In measuring grape quality, there is also a need for
objective measurement of negative quality parameters such
as the degree of mould contamination, particularly with
mechanically harvested grapes, where visual assessment
can be diffi cult. Assessment of grapes for fungal infection
at the weighbridge would normally be done by visual
inspection, but this can be difficult with mechanically
harvested fruit. In this context, the use of vis-NIR spec-
troscopy was reported for the detection of powdery mil-
dew (Erysiphe necator) in wine grapes.
24,29
Samples of
Chardonnay grapes with varying degrees of powdery mil-
dew infection (classifi ed visually) were homogenised then
500 1000 1500 2000 2500
0.5
1.0
1.5
2.0
2.5
Whole grape
Homog. grape
Single berry
Figure 1. Visible and near infrared refl ectance mean spectra of
a single berry, whole grape bunch and homogenate red grape
samples.
282 Review: Analysis of Grapes and Wine by NIR
scanned in refl ectance mode over a 400–2500 nm wave-
length range. The homogenates were also analysed for
powdery mildew DNA content.
29
Powdery mildew DNA
content correlated with the visual infection classifi cation
and strong spectral correlations with infection level were
also observed, including spectral changes not related to pH
and TSS differences. There were no systematic inter-cor-
relations of infection level with pH and TSS, precluding
the possibility that the infection level calibrations could be
based on these parameters. The spectral data were reduced
by principal component analysis (PCA) and the fi rst four
PCA scores were used with a quadratic discriminant analy-
sis algorithm to classify infection level. A 100% classifi -
cation rate was achieved in calibration mode and a 92%
classifi cation rate was achieved using cross-validation. The
PLS regression was used to predict the powdery mildew
DNA content using the spectral data where the degree of
accuracy was suffi cient to clearly discriminate the lowest
infection level (1–10%) from the uninfected samples. Only
two PLS factors were used, indicating the strength of the
spectral correlations, but it must be noted that this was a
small data set (n = 25) and this work must be confi rmed
with larger, more diverse sample sets. The implication
of this work is that it might be possible to discriminate
infected fruit at the weighbridge to provide a “go/no-go”
test to highlight suspect fruit for further detailed analysis to
determine suitability for winemaking.
29
Measurement of wine composition
Since the early work of Kaffka and Norris
4
describing the
use of NIR spectroscopy to measure the alcohol and sugar
content in wine, not many applications have been reported
on the analysis of wine by NIR until recently. Most of the
NIR applications in the wine industry have concentrated
on the measurement of alcohol content. There are a number
of commercially available dedicated NIR-based alcohol
analysers and this technique has now become a standard
analysis method worldwide in the wine industry.
5,7,11
It is
well known that ethanol has a strong NIR absorbance signal
in alcoholic beverages, with an intensity second only to
water; however, accuracy and robustness of the calibrations
obtained can be limited by matrix variations, particularly
variations in sugar concentration.
7
The use of NIR spectroscopy has been reported to meas-
ure several wine compositional parameters such as alcohol
content, pH, volatile acidity, organic acids, malic, tartaric,
Reference Parameter Number of
samples
Varieties Wavelength
range (nm)
Range
SECV/
SEP
RPD
Urbano-Cuadrado
et al. 2004
30
Alcoholic degree (%, v/v)
Total acidity (meq L
–1
)
pH
Glycerol (g L
–1
)
Reducing sugars (g L
–1
)
Total sulphur dioxide (mg L
–1
)
24 Red, rose and
white wines
400–2500 10–15
3.5–8.7
3.2–4.0
2–12.4
0.6–9.7
16–149
0.24
0.48
0.07
0.72
0.33
23.5
5.7
2.27
2.4
4.0
10.3
1.8
Sauvage et al.
2002
32
Sodium (mg L
–1
)
Potassium (mg L
–1
)
Magnesium (mg L
–1
)
Calcium (mg L
–1
)
24 White wine 400–2500 5–118
265–1100
78–718
3 0–120
9.22
79.0
14.5
8.09
2.8
2.8
2.7
2.8
Manley et al.
2001
37
Sugar (°Brix)
Malic acid (g L
–1
)
Lactic acid (g L
–1
)
Ethyl carbamate (µg kg
–1
)
97 White varieties
(must and wine)
1000–2500 17–27
0–4.8
0–5.6
0.4–19.3
0.31
1.02
1.34
3.6
5.95
1.2
0.96
1.06
Cozzolino et al.
2004
50
Malvidin 3 glusoside (mgL
–1
)
Pigmented Polymers (mg L
–1
)
Tannins (mg L
–1
)
> 350 Red varieties
(must and wine)
400–2500 14–427
4–103
12–991
28
5.9
131.2
3.5
3.1
1.8
Dambergs et al.
2002
54
Methanol (g L
–1
) 123 Grape spirit 1200–2450 0.02–20.1
0.06 12.5
SEP: standard error of prediction
SECV: standard error of cross validation
RPD: SD/SEP
SD: standard deviation
Table 2. Analysis of grape juice, must and wine by visible and near infrared spectroscopy in transmission mode by various authors.
D. Cozzolino et al., J. Near Infrared Spectrosc. 14, 279–289 (2006) 283
and lactic acids, reducing sugars and sulphur dioxide in
a set of red, rose and white wines (see Table 2).
30,31
Only
accurate NIR calibrations were obtained for the determina-
tion of alcohol (R
2
= 0.98, SEP = 0.24% v / v), pH (R
2
= 0.81,
SEP = 0.07), reducing sugars (R
2
= 0.71, SEP = 0.33 g L
–1
) and
lactic acid (R
2
= 0.81, SEP = 0.41 g L
–1
).
30,31
The same authors
compared the use of NIR and Fourier transform (FT) MIR
spectroscopy to measure several wine parameters.
30,31
The
authors concluded that, for most of the wine compositional
parameters measured, the NIR calibration results were better
than those obtained by FT-MIR, mainly because of the high
signal-to-noise ratio of the MIR method.
30
The same study
evaluated the combination of NIR and MIR spectra yielding
a better calibration model for the wine compositional param-
eters analysed.
30
The use of NIR spectroscopy has also been reported
for the determination of the concentration of sodium (Na),
potassium (K), magnesium (Mg), calcium (Ca), iron (Fe)
and copper (Cu) in white wines.
32
The authors stated that,
although the calibration might be used to predict those
minerals in white wine, caution must be taken due to the low
number of samples used to develop the calibration models
(n = 24) (Table 2).
Sweet wines made from botrytised grapes represent a
complex matrix in that they have very high sugar, acid and
glycerol content in comparison with standard wines.
33
NIR
spectroscopy was explored in combination with a number
of multivariate calibration routines such as MLR, step-
wise regression (SWR), principal components regression
(PCR) and PLS regression, to analyse ethanol, glycerol,
glucose and fructose in botrytis-affected style wines.
33
Both PLS and SWR gave the best performance, in terms
of the lowest SEP relative to mean values, but glycerol
and glucose had high percentage errors with all calibration
routines.
33
The predictive ability of the models expressed
as a percentage of the root mean of the SEP (RMSEP) were
0.87 and 16.54 for ethanol and glycerol using SWR and
1.2. and 17.27 for ethanol and glycerol using PLS as the
regression model.
33
It is well known that water tends to dominate the wine
NIR spectra and may obscure minor wine components.
Figure 2 shows the vis and NIR raw spectra of water and
red wine samples analysed in transmission (1 mm path
length). Two large absorption bands were observed cor-
responding to the O–H bonds around 1400 and 1900 nm
(water and ethanol), respectively. To overcome this issue,
the use of dry extracts as sample preparation [dry extract
spectroscopy by infrared absorption (DESIR)] was exam-
ined to measure total phenolics and sugar content in forti-
fi ed wines (Porto wine).
8
The SEP obtained were 3.14 g L
–1
,
5.92 g L
–1
and 1.30 for D- glucose, D-fructose and total
phenolics, respectively.
Dedicated FT-MIR instruments are now available and
are used extensively in routine analysis of wine by the
industry.
34–36,38–40
The use of FT-MIR spectroscopy has been
proposed and implemented by several research groups for
routine analysis of a large number of wine compositional
parameters such as alcohol content, volatile acidity, pH, tar-
taric acid, lactic acid, glucose plus fructose, acetic acid and
polyphenols.
34–36,38,39
The application of FT-MIR in wine
analysis was of special interest for some applications due to
the presence of sharp and specifi c absorption bands for the
wine constituents.
34–36
The use of NIR spectroscopy was also
evaluated to build calibrations for free amino nitrogen (FAN)
in grape must and malolactic fermentation status of wines.
37
Although calibrations could not accurately quantify the con-
centration of the compounds of interest (malic acid, lactic
acid, FAN), classifi cation models based on soft independ-
ent modelling by class analogy (SIMCA) as a discriminant
method could distinguish between groups of high, medium
and low concentration. The classification rates obtained
ranged between 80 to 88%.
37
Monitoring wine fermentation
Wine fermentation is a complex process in which grape
juice is transformed by microbial action into a high value
product: wine. The modern wine industry needs both fast
and reliable process quality control methods and techniques
in real time in order to assure the quality of the fi nal prod-
uct to the consumer.
41,42
Control of the wine fermentation
process is a very important step during wine production in
order to monitor accurately and rapidly control both sub-
strate (for example, sugars, ethanol, phenolic compounds)
and product quality.
43
Several authors have reported the use
of NIR and fi bre-optics to monitor alcoholic fermentation
in a diverse group of beverages such as table wine, forti-
fi ed wine, champagne and beer.
40–50
However, according
to these authors, accuracy was limited due to hardware
restrictions. For example, the use of telecommunication
500 1000 1500 2000 2500
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
Log 1/T
Wavelength (nm)
Water
RedWine
Figure 2. Visible and near infrared raw spectra of water and red
wine samples analysed in transmission mode using 1 mm path
length.
284 Review: Analysis of Grapes and Wine by NIR
grade fi bre-optics result in a spectrum with high intrinsic
absorbance and high noise in critical wavelengths used to
predict compositional parameters in wine. Noiseux and co-
workers
44
tested a combined fi bre-optic air-gap absorption
sensor as a tool for wine process monitoring. The combi-
nation of both sensors was used successfully to measure
colour density and refractive index in wines of different
grape varieties.
44
The goal of process control by NIR spectroscopy for wine
fermentation has also been pursued.
45,47
One study described
a fermenter sampling system, with temperature equilibra-
tion, linked to a filter-based, transflectance instrument.
46
A qualitative analysis of the spectra of sugar and ethanol
showed the infl uence of water which masks an important
sugar absorption around 2080 nm.
46
Wavelength selection
was confi rmed by examining spectra from a scanning instru-
ment and calibrations suitable for determining sugar (glucose
plus fructose) and ethanol in fermenting grape must were
developed using two wavelengths (2139 and 2230 nm).
46,47
Comparison of a fi lter-based NIR instrument with a FT-NIR
scanning instrument for the same application revealed sim-
ilar accuracy at medium to high sugar levels.
47
The SEC
values obtained were 0.15% and 2.6 g L
–1
for ethanol and
sugars, respectively.
47
However, the FT-NIR scanning instru-
ment offered better precision at the low sugar levels encoun-
tered near the end of fermentation. Previous authors also
noted that calibrations for sugar determination in individual
ferments had better accuracy than calibrations based on
combined data from different ferments. This calibration
specifi city may be related to the possibility that although the
sample matrix changes during the fermentation process, the
base spectrum is likely to be similar, with the major changes
being in the most abundant constituents of interest, sugar
and ethanol.
47
A fundamental problem with monitoring fer-
mentation is the fact that the sample matrix changes dramati-
cally during the course of fermentation—the use of artifi cial
neural networks (ANN) algorithms has been shown to be
very effective in reducing prediction error compared with
discriminant PLS, using model fermentation systems.
41
This
study has also shown the potential use of NIR spectroscopy
for measuring the kinetics and oenological conditions during
red wine fermentation with great precision.
41
Investigations of process-scale red wine fermentation
trials revealed the potential of vis-NIR spectroscopy to
predict the concentration and monitor the extraction and
evolution of phenolic compounds during red wine fermen-
tation.
41,50
Samples were sourced from fermentation trials
conducted during three vintages using two different types
of fermenters, inoculated with two different types of yeast
(Saccharomyces cerivisiae and bayanus). Results showed
that vis-NIR spectroscopy could predict the concentra-
tion of major anthocyanins such as malvidin-3-glucoside
(R
2
= 0.91 and SECV = 28.0 mg L
–1
), pigmented polymers
(R
2
= 0.87 and SECV = 5.9 mg L
–1
) and tannins (R
2
= 0.83
and SECV = 131.1 mg L
–1
), in Cabernet Sauvignon and
Shiraz wines during fermentation.
50
The use of FT-MIR and fibre-optics to monitor large-
scale wine fermentations has also been reported by various
authors.
43,44
These applications showed promise as a rapid
method; however, according to the researchers more studies
are needed in order to use those techniques on-line in the
winery.
Wine quality grading
The ability to accurately assess wine quality is an impor-
tant part of the wine making process, particularly when
allocating batches of wine to styles determined by con-
sumer requirements.
13
Grape pricing is often determined by
the quality category of the resulting wine—so called “end
use” payment.
51
Wine quality, in terms of sensory char-
acteristics, is normally a subjective measure, performed
by experienced winemakers, wine competition judges or
wine tasting panellists.
13
By nature, such assessments can
be biased by individual preferences and may be subject to
day-to-day variation. An objective quality grading method
would, therefore, be of great assistance in the wine indus-
try.
51
Flavour compounds are often present in concentra-
tions below the detection limit of NIR spectroscopy but the
more abundant organic compounds in the wine matrix offer
potential for objective quality grading by this technique. It
has been demonstrated that wine quality rankings (such as
the score or allocation assigned to wines by sensory panels)
for red and fortifi ed wines could be discriminated by vis-
NIR spectroscopy.
51
Furthermore, in Australia it has been
verifi ed that vis-NIR spectroscopy can predict wine quality
as judged by both commercial wine quality rankings and
wine show scores.
51
Correlations between NIR spectra and
sensory data obtained using wine show samples were less
signifi cant in general, in comparison with the commercial
grading data. The R
2
and SECV obtained using a small set
of samples (n = 20) to predict Tawny port wine score were
0.84 and 0.97, respectively. The commercial samples were
all from one major producer, from one growing area and
were graded immediately—ex-vintage, with minimal oak
treatment.
51
For dry red wines, the best calibrations were
obtained with a class of Pinot Noir—a variety that tends
to be produced in limited areas in Australia and would
represent the least matrix variation. Similar to the win-
emakers quality allocation study, the highest loadings were
observed predominantly in the vis region around 520 nm
(wine pigments).
51
Grading of wine by vis-NIR spectroscopy could provide
a rapid assessment or pre-screening tool to add to the
range of analyses available to winemakers. It could allow
preliminary blend allocation of large numbers of batches
of wines prior to sensory assessment. Winemakers may
be able to develop “profi les” for their blends as in-house
vis-NIR calibrations. Calibrations based on sensory scores
will tend to be diffi cult to obtain due to variation between
individual wine tasters and may not pick up compounds
D. Cozzolino et al., J. Near Infrared Spectrosc. 14, 279–289 (2006) 285
that are present at low concentrations, yet have strong sen-
sory properties. Nevertheless, interpretation of spectral
data may provide valuable insight into the more abun-
dant parameters affecting wine quality and highlight the
interactions that occur within the complex wine matrix in
governing sensory properties.
With regard to white wines, commercially available
bottles of Australian Riesling and unwooded Chardonnay
were sourced from a broader wine fl avour study and assessed
by a trained sensory panel for the attributes honey, estery,
lemon, caramel, toasty, perfumed fl oral and passionfruit
aroma properties and overall flavour and sweetness pal-
ate properties. These samples were scanned using vis-NIR
spectroscopy.
52
PLS calibration models between sensory
scores for the attributes and vis-NIR spectra using different
wavelength regions were developed. The results showed
good correlation between spectra and sensory properties
(R > 0.70) for estery, honey, toasty, caramel, perfumed fl o-
ral and lemon, while poor correlations (R < about 0.55)
were found in most cases for passionfruit, sweetness and
overall fl avour, respectively. However, in developing such
models, only a limited number of wines were used (n = 40)
and, therefore, caution must be considered in extending the
applicability of the technique until further validation work is
completed.
51,52
FT-MIR has also been used to explore the possibility of
grading wine samples from the Qualifi ed Denomination of
Origin (QDO) “Rioja”.
53
According to these authors, the
results showed that the calibration procedures using spectra
were adequate to quantitatively classify wine samples from
QDO and to qualitatively distinguish between “adequate”
and “abnormal” wine samples.
Measurement of methanol and ethanol
in grape distillates
Grape spirit is produced by distillation of wine or wine/
rape derived process waste and is used in the production
of fortifi ed wines. Methanol concentrations in grape marc,
one of the major sources of distillation raw materials, can
be high due to the action of mould and bacteria in the raw
product.
54
The methanol concentration in the fi nal product
must be minimised to comply with food regulations and
operating continuous stills can be diffi cult without rapid
methanol analysis to allow fi ne tuning of the stills in a timely
manner.
54
In comparison to wine, the distillation process
streams represent relatively simple matrices, consisting
of pre dominantly ethanol, water and minor quantities of
other volatile organic compounds. Two key compounds that
are routinely monitored during the distillation process are
ethanol and methanol, which have characteristic NIR s pectra
based on differences in relative concentrations of CH
3
groups,
wavelength shifts for OH groups and a CH
2
group unique to
ethanol. NIR calibrations have been developed using both
PLS and MLR methods with transmission spectra of wine
fortifying spirit using gas chromatography (GC) as the refer-
ence method. The PLS calibrations approached the accuracy
of the reference methods, with an R
2
of 0.998 and a SECV of
0.06 g L
–1
for methanol and an R
2
of 0.96 and SECV of 0.08%
v / v for ethanol. The calibrations were very robust, as indi-
cated by high values for the ratio of the standard deviation of
the reference data to the standard error of prediction of the
calibration (RPD > 12). MLR calibrations were less accurate,
but they were robust across vintages.
54
Product authenticity
Verifi cation of authenticity of food in general, and wine in
particular, has become a potential application of NIR spec-
troscopy. Adulteration can take many forms, including the
addition of sugars, acids, volatile oils, over-dilution of con-
centrate, addition of juices of other fruits, use of concentrate
in a “fresh” product and use of low-quality products recovered
from what are normally waste products of manufacture.
55,56
Food adulteration has been practiced since ancient times but
has become more sophisticated in the recent past.
57
Foods or
ingredients most likely to be targets for adulteration include
those which are of high value or are subject to the vagaries
of weather during their growth or harvesting. The practice of
adulteration commonly arises for two main reasons: fi rst, it can
be profi table and second, adulterants can easily be mixed and
are subsequently diffi cult to detect. To counter this problem,
manufacturers subject their raw material and by-products to
a series of quality controls which includes high-performance
liquid chromatography (HPLC), thin layer chromatography
(TLC), enzymatic tests and physical tests, to establish their
authenticity and hence guarantee the quality of the products
manufactured for the consumer.
57–60
Although there are numerous reports on the use of NIR
spectroscopy to authenticate foods, not many published
works were found in the literature on the use of NIR spec-
troscopy to authenticate grape or wine samples. The use of
NIR spectroscopy was reported to discriminate three red
grape varieties, namely Merlot, Tempranillo and Grenache
grown in Spain.
61
In this study, 100% discrimination was
achieved by PCA using cross-validation. Varietal discrimi-
nation was also achieved using two white varieties, namely
Viura and Chardonnay.
25
The use of vis-NIR was also inves-
tigated to discriminate two Australian white wine varieties,
namely Riesling and unwooded Chardonnay, with accuracy
of up to 95%.
57
Note that, in this study, the use of both the
vis and NIR wavelength regions gave discrimination models
with the best calibration statistics.
57
Yeast identifi cation
Yeast identifi cation is an important issue in process-scale
(industrial) fermentations, where contamination with wild
286 Review: Analysis of Grapes and Wine by NIR
strains might introduce undesirable traits. In the last few
years, both microbial and plant metabolite analysis has
shifted from specifi c assays toward broad spectrum methods,
offering both high accuracy and sensitivity in highly complex
mixtures of compounds.
62–64
Both NIR and FT-NIR spectroscopy have been examined
for suitability as tools for yeast and bacterial strain
identifi cation, yeast protein measurement, yeast trehalose
and glycogen measurement and determination of yeast
concentration.
62–69
As well as assessing yeast in various
growth stages, and after heat shock damage, one study
examined the possibility of discrimination between yeast
strains with NIR spectra of yeast slurries, prepared from
yeast grown in synthetic media. Differences in protein
profi les could be seen with electrophoresis methods and this
was correlated with second derivative refl ectance spectra at
longer NIR wavelengths (2000–2500 nm). Sample sets
were limited in size, but PCA plots clearly discriminated
yeast growth phase yeast strain and could detect 10% cross-
contamination of one strain with another as well as heat-
shock induced artifi cally.
62
The potential of NIR spectroscopy as a rapid screening
technique was also demonstrated in order to discriminate
different yeast strains with particular metabolic profi les.
63,65
Ferment supernatants of different Saccharomyces cerevisiae
deletion strains were analysed in the vis-NIR range (400 to
2500 nm) in transmittance mode with a 1 mm pathlength.
9
Multivariate classifi cation models were developed using
the spectra of the yeast strains. PCA models for each yeast
deletion strain were prepared and used to develop the
SIMCA classification models. The models showed that
the deletion strains were correctly classifi ed as different
from the wild-type laboratory strain. This demonstrated the
potential of combining NIR spectroscopy and multivariate
techniques to enable the rapid selection of yeast strains
with specifi c abilities from a large population and classify
samples that have similar characteristics. The combination
of different analytical techniques with multivariate methods
could be used as a tool for fi ngerprinting of yeast strains on
a large scale.
9
Vine tissue analysis
In addition to grape and wine analysis, the use of NIR
spectroscopy has been extended to other applications
such as soil and plant analysis. The prediction of nutrients
in grapevine petioles as a tool for vineyard fertilisation
management has been reported.
70,71
Nitrogen (r = 0.997 and
SEC = 0.068%), potassium (r = 0.993 and SEC = 0.108%)
and phosphorus (r = 0.996 and SEC = 0.021%) were
predicted by NIR spectroscopy in grape petioles which had
been dried and ground; however, no validation was carried
out.
70,71
The authors concluded that uniform procedures are
needed for sampling and for interpretation of the results
obtained.
Future directions
During the past 30 years, an increasing number of
applications has been investigated and used in viticultural
and oenological studies. However, the use of NIR
spectroscopy in the wine industry is still in its infancy.
From the analysis of the scattered information available,
it seems that NIR spectroscopy has been applied at many
stages during wine production. Figure 3 shows the actual
and possible applications of NIR spectroscopy in the grape
and wine industry. For example, red and white grapes are
analysed by NIR for total anthocyanins (colour), sugar
(TSS), pH and acidity, with acceptable SEP to be used
in routine analysis. Wine fermentation can be monitored
using fi bre-optics, at-line or attenuated total refl ectance
(ATR) cells and yeast strains can be identifi ed using NIR.
Wine is analysed routinely for alcohol content using NIR
fi lter instruments and new applications are investigated;
prediction of sensory scores and varietal authenticity are
just some examples.
Undoubtedly, NIR spectroscopy has become a relevant
technique for the grape and wine industry which will
provide a cheap and rapid method of analysis. Further,
the technology offers the exciting prospect of potentially
Figure 3. Current and potential applications of NIR spectroscopy
in the wine industry.
D. Cozzolino et al., J. Near Infrared Spectrosc. 14, 279–289 (2006) 287
providing for the development of small-scale, inexpensive,
portable hand-held instruments, which would be of great
benefi t to the whole supply chain of the industry. Grape
growers and winemakers could benefi t if the compositional
quality of grapes could be rapidly and non-destructively
assessed using spectroscopy at the weighbridge or even
whilst still on the vine. As the technology of spectroscopic
instrumentation and chemometrics advances further, the
resulting spin-offs may further assist the industry in its
quest to defi ne and objectively measure grape and wine
quality and to assure consumers of the quality of the fi nal
product to be enjoyed.
Acknowledgements
The authors wish to thank all of their industry
collaborators and partners, especially the Hardy Wine
Company (Audrey Lim, Chris Bevin and Peter Dawson)
and Orlando Wines (Inca Pearce and Russell Johnstone).
The critical input and continual encouragement from
Professor P.B. Høj, former Director of the Australian Wine
Research Institute, is also acknowledged. Staff at AWRI
who have contributed to the work reported here is also
acknowledged. This project is supported by Australia’s
grapegrowers and winemakers through their investment
body, the Grape and Wine Research and Development
Corporation, with matching funds from the Australian
government and by the Commonwealth Cooperative
Research Centres Programme. The work was conducted by
The Australian Wine Research Institute and forms part of
the research portfolio of the Cooperative Research Centre
for Viticulture.
References
1. B. Rankine, A Manual of Winemaking Practice for
Australia and New Zealand. Pan Macmillan, Sydney,
Australia (1989).
2. Australian Bureau of Statistics, Australian wine and
grape industry (2003).
3. M. Gishen, R.G. Dambergs, A. Kambouris, M.
Kwiatkowski, W.U. Cynkar, P.B. Høj and I.L. Francis,
in Proceedings of 9
th
International Conference on Near
Infrared Spectroscopy, Ed by A.M.C. Davies and R.
Giangiacomo. NIR Publications, Chichester, UK, p. 917
(2000).
4. K.J. Kaffka and K.H. Norris, Acta Aliment. 5, 267
(1976).
5. G.F. Baumgarten, South Afr. J. Enol. Vitic. 8, 75
(1987).
6. G. Kemeny, T. Pokorny, K. Forizs and L. Leko,
Borgazdasag. 31, 127 (1983).
7. B.G. Osborne, T. Fearn and P.H. Hindle, Practical NIR
Spectroscopy with applications in Food and Beverage
Analysis. Longman Scientifi c and Technical, Harlow,
UK (1993).
8. R. Medrano, S.H. Yan, M. Madoux, V. Baeten and
M. Meurens, in Leaping Ahead with Near Infrared
Spectroscopy. Ed by G.D. Batten, P.C. Flinn, L.A. Welsh
and A.B. Blakeney. Royal Australian Chemical Institute,
Victoria, Australia, p. 303 (1995).
9. M. Gishen, R.G. Dambergs and D. Cozzolino, Aust. J.
Grape Wine Res. 11, 296 (2005).
10. R.G. Dambergs, D. Cozzolino, W.U Cynkar, A.
Kambouris, I.L. Francis, P.B. Høj and M. Gishen, Aust.
Grapegrower and Winemaker J. 476, 69 (2003).
11. M. Gishen, and R.G. Dambergs, Aust. Grapegrower and
Winemaker J. 414, 43 (1998).
12. M. Gishen, P.G. Iland, R.G. Dambergs, M.B. Esler, I.L
Francis, A. Kambouris, R.S. Johnstone and P.B. Høj, in
Proceedings 11
th
Australian Wine Industry Technical
Conference, Ed by R. Blair, P. Williams and P.B. Høj.
Winetitles, Adelaide, South Australia, p. 188 (2002).
13. I.L. Francis, P.B. Høj, R.G. Dambergs, M. Gishen, M.A. de
Barros Lopes, I.S. Pretorius, P.W. Godden, P.A.Henschke,
M.J. Herderich and E.J. Waters, in Proceedings 12
th
Australian Wine Industry Technical Conference, Ed
by R. Blair, P. Williams and I.S. Pretorious. Winetitles,
Adelaide, South Australia, p. 85 (2005).
14. R.G. Dambergs, A. Kambouris, M. Gishen and I.L.
Francis, in Proceedings of Modern Viticulture— meeting
market expectations. Winetitles, Adelaide, South
Australia, p. 45 (1999).
15. C. Jaren, J.C. Ortuño, S. Arazuri, J.I. Arana, and M.C.
Salvadores, Int. J. Infrared Milli. Wavs. 22, 1521 (2001).
16. J. Herrera, A. Guesalaga and E. Agosin, Meas. Sci. Tech.
14, 689 (2003).
17. R.G. Dambergs, D. Cozzolino, W.U. Cynkar, M.B. Esler,
L.J. Janik, I.L. Francis and M. Gishen, in Proceedings of
11
th
International NIR Conference. Ed by A.M.C. Davies
and A. Garrido-Varo. NIR Publications; Chichester, UK,
p. 183 (2003).
18. P.C. Singh, S. Bhamidipati, R.K. Singh, R.S. Smith and
P.E. Nelson, Food Control 7, 141 (1996).
19. A.M. Kennedy, in Proceedings 11
th
Australian Wine
Industry Technical Conference, Ed by R. Blair, P.
Williams and P.B. Høj. Winetitles, Adelaide, South
Australia, p. 199 (2002).
20. M. Gishen and R.G. Dambergs, NIR news 13, 5 (2002).
21. M. Gishen and R.G. Dambergs, NIR news 14, 26
(2002).
22. M.B. Esler, M. Gishen, I.L. Francis, R.G. Dambergs,
A. Kambouris, W.U. Cynkar and D.R. Boehm, in
Near Infrared Spectroscopy: Proceedings of the 10
th
International Conference, Ed by R.K. Cho and A.M.C.
Davies. NIR Publications, Chichester, UK, p. 249
(2002).
23. D. Cozzolino, M.B. Esler, R.G. Dambergs, W.U.
Cynkar, D.R. Boehm, I.L. Francis and M. Gishen, J.
Near Infrared Spectrosc. 12, 105 (2004).
288 Review: Analysis of Grapes and Wine by NIR
24. R.G. Dambergs, B. Stummer, T. Zanker, D. Cozzolino,
M. Gishen and E. Scott, in Proceedings 12
th
Australian
Wine Industry Technical Conference, Ed by R. Blair,
P. Williams and I.S. Pretorious. Winetitles, Adelaide,
South Australia, p. 333 (2005).
25. C. Arana, C. Jaren and S, Arazuri, J. Near Infrared
Spectrosc. 13, 349 (2005).
26. F. Chauchard, R. Codgill, S. Roussel, J.M. Roger and
V. Bellon-Meurel, Chemometr. Intell. Lab. Syst. 71, 141
(2004).
27. J.S. Shenk, P. Berzaghi and M.O. Westerhaus, J. Near
Infrared Spectrosc. 5, 223 (1997).
28. D. Cozzolino, W.U. Cynkar, L.J. Janik, R.G. Dambergs,
I.L. Francis and M. Gishen, in Proceedings 12
th
Australian Wine Industry Technical Conference, Ed by
R. Blair, P. Williams and I.S. Pretorious. Winetitles,
Adelaide, South Australia, p. 335 (2005).
29. B.E. Stummer and E. Scott, in Proceedings 11
th
Australian Wine Industry Technical Conference, Ed by
R. Blair, P. Williams and P.B. Høj. Winetitles, Adelaide,
South Australia, p. 259 (2002).
30. M. Urbano-Cuadrado, M.D. Luque de Castro, P.M.
Perez-Juan, J. Garcia-Olmo and M.A. Gomez-Nieto,
Anal. Chim. Acta. 527, 81 (2004).
31. M. Urbano-Cuadrado, M.D. Luque de Castro, P.M.
Perez-Juan and M.A. Gomez-Nieto, Talanta 66, 218
(2005).
32. L. Sauvage, D. Frank, J. Setarne and M.B. Millikan,
Anal. Chim. Acta 458, 223 (2002).
33. C.M. Garcia-Jares and B. Medina. Fresen. J. Anal.
Chem. 357, 86 (1997).
34. D. Picque, T. Cattenoz, G. Corrieu and J.-L. Berger, Sci.
Alimen. 25, 207 (2005).
35. C.-D. Patz, A. Blieke, R. Ristow and P. Dietrich. Anal.
Chim. Acta 513, 81 (2004).
36. C.-D. Patz, A. David, K. Thente, P. Kurbelm and P.
Dietrich, Vitic. Enol. Sci. 54, 80 (1999).
37. M. Manley, A. van Zyl and E.E.H. Wolf, South Afr. J.
Enol. Vitic. 22, 93 (2001).
38. H.H. Nieuwoudt, B.A. Prior, I.S. Pretorius, M. Manley
and F.F. Bauer, J. Agric. Food Chem. 52, 3726 (2004).
39. M. Gishen and M.G. Holdstock, Aust. Grapegrower and
Winemaker J. 438, 75 and 81 (2000).
40. R.G. Dambergs, M.B. Esler and M. Gishen, in Near
infrared spectroscopy in agriculture. Ed by C.A.
Roberts, J. Workman and J.B. Reeves, III, Agronomy
Monograph 44. ASA, CSSA and SSSA, Madison, WI,
USA. p. 465 (2004).
41. J.M. Roger, J.M. Sablayrolles, J.P Steyer and V. Bellon-
Meurel, Biotech. BioEng. 79, 804 (2002).
42. M. Blanco, A.C. Peinado and C. Mas, Biotech. BioEng.
88, 536 (2004).
43. A. Uturbia, R. Perez-Correa, M. Meurens and E. Agosin,
Talanta 66, 778 (2004).
44. W. Noiseaux, A. Long, A. Cournoyer and M. Vernon,
Appl. Spectrosc. 58, 1010 (2004).
45.
B.R. Buchanan, D.E. Honigs, C.J. Lee and W. Roth,
Appl. Spectrosc. 42, 1106 (1988).
46. J.C. Bouvier, Sci. Aliment. 8, 227 (1988).
47. P. Davenel, B. Grenier, J.C. Foch, P. Bouvier, P. Verlaque
and J. Pourcin. J. Food Sci. 56, 1635 (1991).
48. Y. Li and C.W. Brown, J. Near Infrared Spectrosc. 7,
101 (1999).
49. L.R. Kington and T.M. Jones, in Handbook of Near-
Infrared Analysis, Ed by D.A. Burns and E.W. Ciurczak.
Marcel Dekker: New York, USA, p. 535 (2001).
50. D. Cozzolino, M.J. Kwiatkowski, M. Parker, W.U.
Cynkar, R.G. Dambergs, M. Gishen and M. Herderich,
Anal. Chim. Acta 513, 73 (2004).
51. R.G. Dambergs, A. Kambouris, N. Schumacher, I.L.
Francis, M.B. Esler and M. Gishen, in Proceedings
of the
10
th
International Near Infrared Spectroscopy
Conference, Ed by R.K. Cho and A.M.C. Davies. NIR
Publications, Chichester, UK, p. 187 (2001).
52. D. Cozzolino, H.E. Smyth, K.A. Lattey, W.U. Cynkar, L.
Janik, R.G. Dambergs, I.L. Francis and M. Gishen, Anal.
Chim. Acta 539, 341 (2005).
53. R. Lleti, E. Melendez, M.C. Ortiz, L.A. Sarabia and
M.S. Sanchez, Anal. Chim. Acta 544, 60 (2005).
54. R.G. Dambergs, A. Kambouris, I.L. Francis and M.
Gishen, J. Agric. Food Chem. 50, 3079 (2002).
55. G. Downey, J. Near Infrared Spectrosc. 4, 47 (1996).
56. Ch. Cordella, I. Moussa, A.-C. Martel, N. Sbirrazzuoli
and L. Lizzani-Cuvelier, J. Agric. Food Chem. 50, 1751
(2002).
57. D. Cozzolino, H.E. Smyth and M. Gishen, J. Agric.
Food Chem. 52, 7701 (2003).
58. T. Sato, S. Kawano and M. Iwamoto, J. Dairy Sci. 73,
3408 (1990).
59. K. Thyoldt and T. Isaksson, J. Sci. Food Agric. 73, 525
(1997).
60. P.R. Arhurst and M.J. Dennis, Food Authentication.
Chapman-Hall, London, UK (1996).
61. C. Jaren, C. Olangua, S. Arazuri and J.I. Arana, “Varieties
discrimination in grapes using NIR technologies”,
AgEng2002, p. 235 (2002).
62. A. Halasz, A. Hassan, A. Toth and M. Varadi, Z. Lebensm.
Unters. Forrsch A. 204, 72 (1997).
63. L.J. Sweetlove, R.L. Last and A.R. Fernie, Plant Physiol.
132, 420 (2004).
64. M. Majara, F.M. Mochaba, E.S.C. O’Connor-Cox,
B.C. Axcell and A. Alexander, J. Inst. Brew. 104, 143
(1998).
65. M. Kansiz, P. Heraud, B. Wood, F. Burden, J. Beardall,
and D. McNaughton, Phytochemistry 52, 407 (1999).
66. L.E. Rodriguez-Saona, F.M. Khambaty, F.S. Fry and
E.M. Calvey, J. Agr. Food Chem. 49, 574 (2001).
67. F. Mochaba, P. Torline and B. Axel, J. Am. Soc. Brew.
Chem. 52, 145 (1994).
68. N. Moonsamy, F. Mochaba, M. Majara, E.S.C. O’Connor-
Cox and B.D. Axcell, J. Inst. Brew. 101, 203 (1995).
69. A.M. Upperton, J. Inst. Brew. 75, 464 (1995).
D. Cozzolino et al., J. Near Infrared Spectrosc. 14, 279–289 (2006) 289
70. R.J. Hutton, P.J. Sinclair, A.B. Blakeney and G.D. Batten,
in Leaping Ahead with Near Infrared Spectroscopy,
Ed by G.D. Batten, P.C. Flinn, L.A. Welsh and A.B.
Blakeney. Royal Australian Chemical Institute, Victoria,
Australia, p. 256 (1994).
71. C.R. Blatt, G.D. Batten and A.B. Blakeney, in Near
infrared spectroscopy: the future waves, Ed by A.M.C.
Davies and P. Williams. NIR Publications, Chichester,
UK, p. 107 (1996).
Received: 16 March 2006
Revised: 28 September 2006
Accepted: 29 September 2006
Web Publication: 21 November 2006
