Research Article Received: 2 October 2014
Revised: 11 December 2014
Accepted article published: 18 December 2014
Published online in Wiley Online Library: 14 January 2015
(wileyonlinelibrary.com) DOI 10.1002/jsfa.7053
Rapid monitoring of grape withering using visible near-infrared spectroscopy Roberto Beghi,* Valentina Giovenzana, Simone Marai and Riccardo Guidetti Abstract BACKGROUND: Wineries need new practical and quick instruments, non-destructive and able to quantitatively evaluate during withering the parameters that impact product quality. The aim of the work was to test an optical portable system (visible near-infrared (NIR) spectrophotometer) in a wavelength range of 400–1000 nm for the prediction of quality parameters of grape berries during withering. RESULTS: A total of 300 red grape samples (Vitis vinifera L., Corvina cultivar) harvested in vintage year 2012 from the Valpolicella area (Verona, Italy) were analyzed. Qualitative (principal component analysis, PCA) and quantitative (partial least squares regression algorithm, PLS) evaluations were performed on grape spectra. PCA showed a clear sample grouping for the different withering stages. PLS models gave encouraging predictive capabilities for soluble solids content (R2 val = 0.62 and ratio performance deviation, RPD = 1.87) and firmness (R2 val = 0.56 and RPD = 1.79). CONCLUSION: The work demonstrated the applicability of visible NIR spectroscopy as a rapid technique for the analysis of grape quality directly in barns, during withering. The sector could be provided with simple and inexpensive optical systems that could be used to monitor the withering degree of grape for better management of the wine production process. © 2014 Society of Chemical Industry Keywords: grape withering; visible NIR spectroscopy; chemometrics; postharvest; soluble solids content; firmness
INTRODUCTION
3144
The postharvest withering of grapes is the key process for the production of passito wines.1 – 3 In Italy, sweet wines (dessert wines) and non-sweet wines (as well as some wines derived from grapes partially dehydrated such as Amarone and Sfurzat in northern Italy) are produced from grapes withered in an open or closed environment. Modification of the drying environment greatly affects grape metabolism, and thus the final quality of the wine.4 During this process, berry dehydration and the development of fungi on the grapes cause several physicochemical changes in juice composition. A limited literature is available regarding the withering process.4 – 7 Drying kinetics depend on experimental conditions like temperature, relative humidity, air flow and initial physicochemical characteristics of the product. Studies were conducted on the modifications in color, volatile compounds and phenolic composition induced in wine grapes by off-vine drying.8 In traditional withering, grapes are placed on trays in a single layer and stored in barns (fruttai) under ambient conditions for approximately 3 months. During withering, grapes undergo a weight loss up to 30% due to water evaporation. Although very few studies have been conducted on the modification of mechanical characteristics during the grape drying process, berry skin texture parameters have already been considered efficient indicators to assess the wine grape suitability for drying.9 These changes may affect the optical properties of the skin of the grape berries during withering. Established methods for quality assessment are generally based on analyses characterized by difficult preparation of samples and requirement of a well-equipped laboratory, with results J Sci Food Agric 2015; 95: 3144–3149
available only after several hours. The lack of readily available data means that the wine-making process must begin without having the necessary information, thus reducing their chances of diversifying production and obtaining products with the desired characteristics. A tool enabling real-time analysis during withering would allow preliminary decisions about the process thanks to the rapid analysis of technological parameters simultaneously. Near-infrared (NIR) spectroscopy has gained wide acceptance in different fields, in particular in postharvest fruit and vegetable production.10 The main advantage of NIR technology, in fact, is its ability to record spectra for solid and liquid samples without any pretreatment. This characteristic makes it attractive for the speedy analysis of products. Cost savings are often achieved for NIR measurements related to improved control and product quality, and the technique can provide results significantly faster compared to traditional laboratory analysis. The success of NIR technology has resulted in a good number of publications dealing with different aspects of NIR technology. Nicolaï et al.11 gave a comprehensive overview of NIR spectroscopy for measuring the quality attributes of fruit and vegetables. In particular, they paid
∗
Correspondence to: Roberto Beghi, Department of Agricultural and Environmental Sciences – Production, Landscape, Agroenergy (DiSAA), Università degli Studi di Milano, via Celoria 2, 20133 Milan, Italy. E-mail: [email protected] Department of Agricultural and Environmental Sciences – Production, Landscape, Agroenergy (DiSAA), Università degli Studi di Milano, 20133 Milan, Italy
www.soci.org
© 2014 Society of Chemical Industry
Monitoring of grape withering using visible NIR
www.soci.org
attention to optics and chemometrics as well as applications, and identified areas where more research is required. Many studies reported in the literature have evaluated the applicability of NIR spectroscopy to obtain a complete check on fruits. For example, several applications to estimate the ripeness parameters of different fruit species can be found, especially on grapes12 – 17 and leaves.18 Moreover, in recent years, interest has shifted towards the development of portable systems which can also be used in orchards and during postharvest. NIR instruments require a reliable hardware system and complex mathematical techniques to explain chemical information encoded in spectral data.19 The most commonly used chemometric techniques are (i) spectral preprocessing, to remove any irrelevant information, (ii) principal component analysis (PCA), to perform qualitative data analysis and (iii) partial least square (PLS) regression analysis, to obtain a quantitative prediction of interesting parameters.11,20 – 22 Only a few studies have been published regarding the application of non-destructive techniques for withering monitoring. An NIR acousto-optic tunable filter device (1100–2300 nm) was tested by Bellincontro et al.23 on Aleatico (Vitis vinifera L.) grapes, dehydrated at 20 ∘ C, 45% relative humidity and 1.5 m s−1 air flow in a small-scale thermo-conditioned tunnel, in order to predict total soluble solids and moisture content. The aim of the work was to test an optical portable system (visible NIR spectrophotometer) in the wavelength range 400–1000 nm for prediction of quality parameters of grape berries during withering (not forced dehydration). The chemometric analysis was focused on the correlation between the visible NIR spectra and the technological parameters most representative and interrelated with the process (firmness and soluble solids content, SSC), in order to validate the effectiveness of the system.
MATERIALS AND METHODS Sampling Sampling was performed six times on cv. Corvina (Vitis vinifera L.) berries stored in an experimental natural drying fruit room (fruttaio, barn) equipped with a monitoring system for environmental conditions, located in the Valpolicella winemaking area (Italy), from 30 September to 1 December 2012. During withering, environmental conditions change. Temperature and relative humidity (RH) start from 16 ∘ C and 65% RH and reach 9 ∘ C and 75% RH after 70 days. The optimal withering conditions were determined from previous studies which allowed us to identify the best combination of temperature and RH during the process, and in particular the best withering kinetics able to ensure high-quality withered grapes. For each sampling date 50 berries were randomly selected from bunches located in different points of the barn (barn dimensions were 500 × 600 cm). A total of 300 red berry samples (Vitis vinifera L. cv. Corvina), were analyzed. On selected berries non-destructive analysis was carried out using a portable commercial visible NIR spectrophotometer (two acquisitions were performed in the equatorial region and then averaged); physicochemical analyses were then performed.
J Sci Food Agric 2015; 95: 3144–3149
reflection measurement; (iii) a spectrophotometer; (iv) hardware for data acquisition and instrument control; (v) a battery as the power supply. Spectra were acquired in reflectance mode, without any sample preparation. Light radiation was guided from the light source to the sample through a Y-shaped, bidirectional fiber optic probe (Ocean Optics) in which seven fibers are arranged in a six-around-one configuration, each with a diameter of 600 μm. The six external fibers guide the light from the light source to the sample, while the single central fiber brings it back from the sample to the spectrophotometer. The tip of the optical probe was equipped with a soft plastic cap to ensure contact with the sample’s skin during measurements, while minimizing environmental light interference (Fig. 1). The integrated spectrophotometer was equipped with diffractive grating for spectral measurements optimized in the range of 400–1000 nm and a CCD sensor with a 2048 pixel matrix, corresponding to a nominal resolution of 0.3 nm. A calibration procedure was designed in order to minimize the effects due to possible changes in the light source or detector. The system requires the compulsory use of dark and white references (White Reflectance Standard by Edmund Optics, Barrington, NJ, USA) to finalize the calibration procedure. Furthermore, to avoid the risk of signal stability problems, the acquisition of five spectra for each measure, immediately averaged, was performed. Physical and chemical analyses The SSC was measured using a digital pocket refractometer (DBX-55 Atago, Tokyo, Japan). A few drops of the juice for each berry were placed on the refractometer sensor, previously calibrated with distilled water. The instrument gave the result directly in ∘ Brix. Firmness was measured on individual fruit using a rapid non-destructive instrument. The firmness test was performed using a Durofel® (CTIFL Copa Technologie, Saint Etienne du Grès, France) dynamometer with a bolt of 3 mm diameter (0.10 cm2 ), on a scale of 1 (soft) to 60 (firm). Results were expressed as Durofel index (DI).24 Weight loss determination was performed using an analytical balance (model PE 6000, Mettler-Toledo, Zurich, Switzerland). Data processing Mean, standard deviation (SD) and confidence intervals (𝛼 = 0.05) were calculated from the SSC, firmness and weight loss data. Chemometric analysis was performed using The Unscrambler®
© 2014 Society of Chemical Industry
wileyonlinelibrary.com/jsfa
3145
Portable visible NIR device Spectral acquisitions were performed on grape berries directly in barns using a commercial visible NIR spectrophotometer (Jaz series, Ocean Optics, Winter Par, FL, USA) operating between 400 and 1000 nm. The Jaz system consists of five components: (i) a visible NIR lighting system (halogen lamp); (ii) a fiber optic probe for
Figure 1. Detail of the fiber optic probe during spectral acquisitions on grape berry with vis/NIR portable system.
www.soci.org 9.8 software package (CAMO ASA, Oslo, Norway). Moving average smoothing (15 nm wide window), first and second derivative treatments were applied to visible NIR spectra, aimed at reducing noise before building the calibration models. Derivatives were performed using the Norris Gap transformation (11 nm wide window). PCA was performed on visible NIR spectra to examine sample grouping in order to discriminate samples and to identify outliers. PCA is a linear and unsupervised procedure that permits useful information to be extracted from the data, to explore the data structure and the relationship between objects and the global correlation of the variables.25 The visible NIR spectra were correlated with quality parameters (SSC and firmness) using the PLS regression algorithm. Models were validated using both internal validation technique (cross-validation) and test set validation with an independent validation set. Cross-validation is an internal validation method in which some samples, in this case only one (leave-one-out cross-validation) are omitted from the calibration and used for validation. This is repeated until all samples have been omitted once. To evaluate model accuracy, the coefficient of determination in calibration (R2 cal ), root mean standard error of calibration (RMSEC), coefficient of determination in validation and cross-validation (R2 val and R2 cv ), root mean standard error of validation and cross-validation (RMSEP and RMSECV) were calculated. Finally, for each elaborated model, the ratio performance deviation (RPD) was calculated. RPD is defined as the ratio between the standard deviation of the response variable and RMSE in validation. An RPD ratio of less than 1.5 indicates incorrect predictions and the model cannot be used for further prediction. An RPD between 1.5 and 2 means that the model can discriminate low from high values of the response variable; a value between 2 and 2.5 indicates that coarse quantitative predictions are possible, and a value between 2.5 and 3 or above corresponds to good and excellent prediction accuracy, respectively.26 – 28 The developed PLS models were also tested also using independent validation sets, randomly created, and comprising 40% of the total samples. Optimum calibrations were selected based on minimizing the RMSEP and the RMSECV, and maximizing the RPD.
RESULTS AND DISCUSSION The average measured raw spectra of the six sampling times during the unforced withering process are shown in Fig. 2, with
3146
Figure 2. Average raw spectra of grapes grouped by different sampling times during withering (from t0 to t5). Vertical bars indicate the standard deviation within each group at different wavelengths.
wileyonlinelibrary.com/jsfa
R Beghi et al.
bars indicating the standard deviation within each group at different wavelengths. As expected, the spectra exhibit significant differences among times, with dramatic changes in the visible NIR bands occurring from time zero (t0) to the last withering stages (t3–t5), which is especially linked to firmness modification with loss of berry water, and consequently loss of weight and turgidity, during withering (Fig. 3). In particular, this leads to a strong decrease in reflectance in the visible band associated with the anthocyanin absorption peak centered on 540 nm. Changes in spectra obviously reflect modifications in quality parameters during withering. Figure 3 also summarizes the evolution of measured parameters (SSC and firmness) during the process, compared with the weight loss trend. SSC and firmness are time dependent; in fact, correlation coefficients of R2 = 0.98 and R2 = 0.96 for SSC and firmness, respectively, were noted. The natural loss of water causes an increase of SSC, from 24.0 ± 0.8 ∘ Brix to 29.9 ± 0.5 ∘ Brix. Regarding firmness, on the contrary, a decrease from 34.8 ± 2.9 DI to 19.0 ± 0.9 D.I. was detectable. In order to evaluate the ability of the visible NIR device to differentiate samples during the withering process, data were elaborated by PCA performed on the spectral matrix. Figure 4 shows that the PCA scores plot in the plane defined by the first two principal components (PC1 and PC2) accounted for 96% of the total variance. A clear distribution of samples along PC1 and PC2 according to the sampling times during withering was found. In particular, the grape spectra of the first four sampling points (from t0 to t3, corresponding to about 40 days of withering) are distributed along the PC1. This component explains 90% of the total variance and, in fact, during the first 40 days the most evident changes to the berries occurred, in accordance with the weight loss trend (Fig. 3). Conversely, the spectra of the samples with sampling times from t3 to t5 (corresponding to the last 30 days of withering) are distributed along PC2, which explains 6% of the total variance. Over the last 30 days, in fact, a slowdown of the withering process is noted, with less evident changes in berry characteristics. Descriptive statistics and the statistics related to the PLS models obtained by visible NIR spectroscopy for analyzed indices (SSC and firmness) during withering are shown in Tables 1 and 2. Average data are shown, considering all the sampling times. Different spectral pretreatments were tested. Table 1 reports PLS models elaborated using internal validation with cross-validation technique, whereas Table 2 shows PLS models elaborated using an independent test set for validation. The models developed for SSC show a fairly good determination coefficient in calibration, ranging from 0.67 to 0.75. In cross-validation R2 ranged from 0.55 to 0.68 and the values of RPD were slightly less than 2 (1.67–1.87). The best model was developed using only the smoothing pretreatment of the spectra (Table 1). Results regarding models elaborated using an independent validation set were similar, with the same RPD values in validation. The model derived from smoothed spectra was confirmed to be the best one (Table 2). The performance of the regression models obtained for SSC prediction showed no particular differences from the results reported in the literature on other fruits.11 Bellincontro et al.23 tested an acousto-optical tunable filter (AOTF)-NIR spectrophotometer (1100–2300 nm) on Aleatico grape during controlled dehydration, obtaining good results for SSC prediction (R2 = 0.93 in cross-validation and R2 = 0.92 using an independent validation set). These results are better than those obtained in the present work. This is mainly related to the presence of the two dominant peaks, broad due to water, near 1440 and 1930 nm in the NIR region.29 The spectral range considered in the present work contain only the weaker third overtone of the OH bond (760 nm). Nevertheless, the choice to use a cheaper portable
© 2014 Society of Chemical Industry
J Sci Food Agric 2015; 95: 3144–3149
Monitoring of grape withering using visible NIR
www.soci.org
Figure 3. Evolution of SSC and firmness compared with the weight loss trend during withering. Equations of the trend lines and R2 for SSC and firmness are shown. Vertical bars indicate the confidence interval (𝛼 = 0.05) for each sampling time.
Figure 4. Scores plot of PC1 versus PC2 of PCA on grape spectra, divided by sampling times during withering process (from t0 to t5).
J Sci Food Agric 2015; 95: 3144–3149
CONCLUSIONS A portable optical device (visible NIR spectrophotometer in the wavelength range 400–1000 nm) for single-sample, non-destructive and rapid prediction of grape quality parameters directly in barns during the withering process was tested. In particular, the study was performed to estimate SSC and firmness of grape berries. The prediction capabilities of the system for both parameters were evaluated through the elaboration of prediction models based on multivariate regression techniques. Preliminary results were encouraging and the robustness of prediction models was satisfactory when tested with an independent validation set, with no significant losses of predictive power. Further studies are needed to confirm these early results, and the use of more accurate instrumentation for the reference data is desirable. Moreover, the investigation of wavelengths in order to highlight and select the most informative bands could improve the prediction capabilities of the models to monitor grapes during withering. The work has
© 2014 Society of Chemical Industry
wileyonlinelibrary.com/jsfa
3147
visible NIR spectrophotometer, compared with a conventional bench-top NIR device, is due to a potential future affordable use of the tested instrument even by small wineries. Similar results to the authors’ findings were obtained by Guidetti et al.30 using a portable visible NIR system (450–980 nm) for fresh berries of red Nebbiolo grapes, with R2 cv = 0.72 and RMSECV = 0.79 ∘ Brix for SSC. Finally, Larrain et al.31 achieved an RMSECV ranging from 1.01 to 1.27 ∘ Brix using a portable NIR device (640–1300 nm) on wine grapes under field conditions. The firmness data showed coefficients of determination in calibration from 0.51 to 0.62 and in cross-validation from 0.51 to 0.60. Fair RPD values were obtained in cross-validation, ranging from 1.64 to 1.71. The best model was developed using the first derivative pretreatment, thanks to a smaller number of latent variables (LV) with respect to the model derived from smoothed spectra (Table 1). Results were partially confirmed using an independent validation set, with R2 cal ranging from 0.64 to 0.72 and R2 val from 0.49 to 0.56. RPD values in validation were similar to those obtained in cross-validation and also in this case the best model was calculated starting from first derivative pretreatment. The analysis of firmness through the
application of visible NIR and NIR spectroscopy often shows difficulties, as highlighted by some published studies.32,33 These difficulties are mainly due to the extreme variability of this parameter among berries, particularly during a process with high variation in textural characteristics like withering. Other factors are the high instrumental error of the portable dynamometer and, also, the difficulty of calibrating a model for the estimation of an index not directly associable with a chemical species (and consequently the absorption bands of those chemical bonds). For these reasons, the predictive capabilities of the model for this parameter are usually poor. In this work, however, an R2 of about 0.6 in validation and RPD slightly less than 2 show the possibility of distinguishing between high and low firmness values.27,28 Similar results were obtained for firmness prediction on intact olives by Beghi et al.34 and by Kavdir et al.,35 using in this case FT-NIR spectroscopy in the wavelength range 780–2500. No results can be found in the literature regarding the application of visible NIR spectroscopy for the prediction of textural features of grapes during the withering process.
www.soci.org
R Beghi et al.
Table 1. Descriptive statistics and statistics of the PLS models elaborated on visible NIR spectra to estimate qualitative parameters of grape during withering, using leave-one-out cross-validation technique Calibration model Quality parameter
Mean
SD
Pretreatment
LV
n
SSC (∘ Brix)
27.8
2.7
Firmness (DI)
25.2
6.8
Smoothing Der1 Der2 Smoothing Der1 Der2
15 11 5 5 4 2
289 286 282 265 263 263
R2
Cross-validation model R2
RMSEC
0.75 0.69 0.67 0.62 0.62 0.51
1.28 1.37 1.39 3.87 3.86 4.12
RMSECV
0.68 0.62 0.55 0.60 0.60 0.51
1.46 1.54 1.63 3.97 3.97 4.15
RPD 1.87 1.77 1.67 1.71 1.71 1.64
Table 2. Descriptive statistics and statistics of the PLS models elaborated on visible NIR spectra to estimate qualitative parameters of grape during withering, using an independent test set (40% of total samples) for validation Calibration model Quality parameter
Mean
SD
Pretreatment
SSC (∘ Brix)
27.8
2.7
Firmness (DI)
25.2
6.8
Smoothing Der1 Der2 Smoothing Der1 Der2
LV 14 10 5 4 3 4
n
R2
171 173 181 146 149 155
0.79 0.67 0.69 0.64 0.66 0.72
demonstrated the applicability of visible NIR spectroscopy as a rapid technique for the analysis of grape quality during withering. This could provide the sector with simple and inexpensive optical systems which can be used to monitor the withering degree of grapes for better management of the wine production process. These instruments could be an effective solution also for small wineries seeking postharvest methods and sorting systems to improve the quality of their production.
ACKNOWLEDGEMENTS The authors would like to acknowledge Masi Agricola SpA in Gargagnago di Valpolicella (Verona, Italy) for their collaboration.
REFERENCES
3148
1 Domizio P and Lencioni L, Vin Santo. Adv Food Nutr Res 63:41–100 (2011). 2 Giordano M, Rolle L, Zeppa G and Gerbi V, Chemical and volatile composition of three sweet white passito wines. J Int Sci Vigne Vin 43:159–170 (2009). 3 Ribereau-Gayon P, Dubourdieu D, Doneche B, and Lonvaud-Funel A, Handbook of Enology: The Microbiology of Wine and Vinifications. Wiley, Chichester (2006). 4 Barbanti D, Mora B, Ferrarini R, Tornielli GB and Cipriani M, Effect of various thermo-hygrometric conditions on the withering kinetics of grapes used for the production of ’Amarone’ and ’Recioto’ wines. J Food Eng 85:350–358 (2008). 5 Muganu M, Bellincontro A, Barnaba FE, Paolocci M, Bignami C, Gambellini G et al., Influence of bunch position in the canopy on berry epicuticular wax during ripening and on weight loss during postharvest dehydration. Am J Enol Vitic 62:91–98 (2011). 6 Costantini V, Bellincontro A, De Santis D, Rotondi R and Mencarelli F, Metabolic changes of Malvasia grapes for wine production during postharvest drying. J Agric Food Chem 54:3334–3340 (2006). 7 Bellincontro A, De Santis D, Botondi R, Villa I and Mencarelli F, Different postharvest dehydration rates affect quality characteristics and volatile compounds of Malvasia, Trebbiano and Sangiovese grapes for wine production. J Sci Food Agric 84:1791–1800 (2004).
wileyonlinelibrary.com/jsfa
RMSEC 1.24 1.45 1.49 3.55 3.59 3.33
Validation model RPD 2.20 1.88 1.83 1.92 1.89 2.04
n
R2
117 117 91 103 101 98
0.62 0.59 0.52 0.56 0.56 0.49
RMSEP 1.46 1.60 1.64 3.87 3.79 4.15
RPD 1.87 1.71 1.66 1.76 1.79 1.64
8 Serratosa MP, Lopez-Toledano A, Merida J and Medina M, Changes in color and phenolic compounds during the raisining of grape cv. Pedro Ximenez. J Agric Food Chem 56:2810–2816 (2008). 9 Rolle L, Torchio F, Giocosa S and Gerbi V, Modifications of mechanical characteristics and phenolic composition in berry skins and seeds of Mondeuse wine grapes throughout the on-vine drying process. J Sci Food Agric 89:1973–1980 (2009). 10 Wang W and Paliwal J, Near-infrared spectroscopy and imaging in food quality and safety. Sens Instrum Food Qual 1:193–207 (2007). 11 Nicolaï BM, Beullens K, Bobelyn E, Peirs A, Saeys W, Theron KI et al., Non-destructive measurement of fruit and vegetable quality by means of NIR spectroscopy: a review. Postharvest Biol Technol 46:99–118 (2007). 12 Lafontaine M, Freund M, Vieth KU and Negara C, Automatic fruit sorting by non-destructive determination of quality parameters using visible/near infrared to improve wine quality. I. Red wine production. NIR News 24:6–8 (2013). 13 Bellincontro A, Nicoletti I, Valentini M, Tomas A, De Santis D, Corradini D et al., Integration of nondestructive techniques with destructive analyses to study postharvest water stress of wine grapes. Am J Enol Vitic 60:57–65 (2009). 14 Fernández-Novales J, López MI, Sánchez MT, Morales J and González-Caballero V, Shortwave-near infrared spectroscopy for determination of reducing sugar content during grape ripening, winemaking, and aging of white and red wines. Food Res Int 42:285–291 (2009). 15 Cozzolino D, Dambergs RG, Janik L, Cynkar WU and Gishen M, Analysis of grapes and wine by near infrared spectroscopy. J Near Infrared Spectrosc 14:279–289 (2006). 16 Gishen M, Dambergs RG and Cozzolino D, Grape and wine analysis – enhancing the power of spectroscopy with chemometrics: a review of some applications in the Australian wine industry. Aust J Grape Wine Res 11:296–305 (2005). 17 Cozzolino D, Esler MB, Dambergs RG, Cynkar WU, Boehm DR, Francis IL et al., Prediction of colour and pH in grapes using a diode array spectrophotometer (400–1100 nm). J Near Infrared Spectrosc 12:105–111 (2004). 18 De Bei R, Cozzolino D, Sullivan W, Cynkar W, Fuentes S, Dambergs R et al., Non-destructive measurement of grapevine water potential using near infrared spectroscopy. Aust J Grape Wine Res 17:62–71 (2011).
© 2014 Society of Chemical Industry
J Sci Food Agric 2015; 95: 3144–3149
Monitoring of grape withering using visible NIR
www.soci.org
19 Cogdill RP and Anderson CA, Efficient spectroscopic calibration using net analyte signal and pure component projection methods. J Near Infrared Spectrosc 13:119–132 (2005). 20 Wold S, Sjöström M and Eriksson L, PLS-regression: a basic tool of chemometrics. Chemom Intell Lab Syst 58:109–130 (2001). 21 Naes T, Isaksson T, Fearn T and Davies T, A User-Friendly Guide to Multivariate Calibration and Classification. NIR Publications, Chichester (2002). 22 Cen H and He Y, Theory and application of near infrared reflectance spectroscopy in determination of food quality. Trends Food Sci Technol 18:72–83 (2007). 23 Bellincontro A, Cozzolino D and Mencarelli F, Application of NIR-AOTF spectroscopy to monitor Aleatico grape dehydration for Passito wine production. Am J Enol Vitic 62:256–260 (2011). 24 Chiabrando V, Giacalone G and Rolle L, Mechanical behaviour and quality traits of highbush blueberry during postharvest storage. J Sci Food Agric 89:989–992 (2009). 25 Beebe KR, Pell RJ and Seasholtz MB, Chemometrics: A Practical Guide. Wiley, New York (1998). 26 Williams PC, Variables affecting near-infrared reflectance spectroscopic analysis, in Near-Infrared Technology in the Agricultural and Food Industries, ed. by Williams PC and Norris KH. American Association of Cereal Chemists, St Paul, MN, pp. 143–167 (1987). 27 Williams PC, Implementation of near-infrared technology, in Near-Infrared Technology in the Agricultural and Food Industries,
28 29 30 31 32
33
34 35
ed. by Williams P and Norris KH. American Association of Cereal Chemists, St Paul, MN, pp. 145–169 (2001). Fearn T, Assessing calibrations: SEP, RPD, RER and R2 . NIR News 13:12–14 (2002). Büning-Pfaue H, Analysis of water in food by near infrared spectroscopy. Food Chem 82:107–115 (2003). Guidetti R, Beghi R and Bodria L, Evaluation of grape quality parameters by a simple Vis/NIR system. Trans ASABE 53:477–484 (2010). Larrain M, Guesalaga AR and Agosin E, A multipurpose portable instrument for determining ripeness in wine grapes using NIR spectroscopy. IEEE Trans Instrum Meas 57:294–302 (2008). Nicolaï BM, Verlinden BE, Desmet M, Saevels S, Theron K, Cubeddu R et al., Time-resolved and continuous wave NIR reflectance spectroscopy to predict soluble solids content and firmness of pear. Postharv Biol Technol 47:68–74 (2008). Zude M, Herold B, Roger JM, Bellon-Maurel V and Landahl S, Non-destructive tests on the prediction of apple fruit flesh firmness and soluble solids content on tree and in shelf life. J Food Eng 77:254–260 (2006). Beghi R, Giovenzana V, Civelli R, Cini E and Guidetti R, Characterization of olive fruit for the milling process by using visible/near infrared spectroscopy. J Agric Eng 44:56–61 (2013). Kavdir I, Buyukcan MB, Lu R, Kocabiyik H and Seker M, Prediction of olive quality using FT-NIR spectroscopy in reflectance and transmittance modes. Biosyst Eng 103:304–312 (2009).
3149
J Sci Food Agric 2015; 95: 3144–3149
© 2014 Society of Chemical Industry
wileyonlinelibrary.com/jsfa
Revised: 11 December 2014
Accepted article published: 18 December 2014
Published online in Wiley Online Library: 14 January 2015
(wileyonlinelibrary.com) DOI 10.1002/jsfa.7053
Rapid monitoring of grape withering using visible near-infrared spectroscopy Roberto Beghi,* Valentina Giovenzana, Simone Marai and Riccardo Guidetti Abstract BACKGROUND: Wineries need new practical and quick instruments, non-destructive and able to quantitatively evaluate during withering the parameters that impact product quality. The aim of the work was to test an optical portable system (visible near-infrared (NIR) spectrophotometer) in a wavelength range of 400–1000 nm for the prediction of quality parameters of grape berries during withering. RESULTS: A total of 300 red grape samples (Vitis vinifera L., Corvina cultivar) harvested in vintage year 2012 from the Valpolicella area (Verona, Italy) were analyzed. Qualitative (principal component analysis, PCA) and quantitative (partial least squares regression algorithm, PLS) evaluations were performed on grape spectra. PCA showed a clear sample grouping for the different withering stages. PLS models gave encouraging predictive capabilities for soluble solids content (R2 val = 0.62 and ratio performance deviation, RPD = 1.87) and firmness (R2 val = 0.56 and RPD = 1.79). CONCLUSION: The work demonstrated the applicability of visible NIR spectroscopy as a rapid technique for the analysis of grape quality directly in barns, during withering. The sector could be provided with simple and inexpensive optical systems that could be used to monitor the withering degree of grape for better management of the wine production process. © 2014 Society of Chemical Industry Keywords: grape withering; visible NIR spectroscopy; chemometrics; postharvest; soluble solids content; firmness
INTRODUCTION
3144
The postharvest withering of grapes is the key process for the production of passito wines.1 – 3 In Italy, sweet wines (dessert wines) and non-sweet wines (as well as some wines derived from grapes partially dehydrated such as Amarone and Sfurzat in northern Italy) are produced from grapes withered in an open or closed environment. Modification of the drying environment greatly affects grape metabolism, and thus the final quality of the wine.4 During this process, berry dehydration and the development of fungi on the grapes cause several physicochemical changes in juice composition. A limited literature is available regarding the withering process.4 – 7 Drying kinetics depend on experimental conditions like temperature, relative humidity, air flow and initial physicochemical characteristics of the product. Studies were conducted on the modifications in color, volatile compounds and phenolic composition induced in wine grapes by off-vine drying.8 In traditional withering, grapes are placed on trays in a single layer and stored in barns (fruttai) under ambient conditions for approximately 3 months. During withering, grapes undergo a weight loss up to 30% due to water evaporation. Although very few studies have been conducted on the modification of mechanical characteristics during the grape drying process, berry skin texture parameters have already been considered efficient indicators to assess the wine grape suitability for drying.9 These changes may affect the optical properties of the skin of the grape berries during withering. Established methods for quality assessment are generally based on analyses characterized by difficult preparation of samples and requirement of a well-equipped laboratory, with results J Sci Food Agric 2015; 95: 3144–3149
available only after several hours. The lack of readily available data means that the wine-making process must begin without having the necessary information, thus reducing their chances of diversifying production and obtaining products with the desired characteristics. A tool enabling real-time analysis during withering would allow preliminary decisions about the process thanks to the rapid analysis of technological parameters simultaneously. Near-infrared (NIR) spectroscopy has gained wide acceptance in different fields, in particular in postharvest fruit and vegetable production.10 The main advantage of NIR technology, in fact, is its ability to record spectra for solid and liquid samples without any pretreatment. This characteristic makes it attractive for the speedy analysis of products. Cost savings are often achieved for NIR measurements related to improved control and product quality, and the technique can provide results significantly faster compared to traditional laboratory analysis. The success of NIR technology has resulted in a good number of publications dealing with different aspects of NIR technology. Nicolaï et al.11 gave a comprehensive overview of NIR spectroscopy for measuring the quality attributes of fruit and vegetables. In particular, they paid
∗
Correspondence to: Roberto Beghi, Department of Agricultural and Environmental Sciences – Production, Landscape, Agroenergy (DiSAA), Università degli Studi di Milano, via Celoria 2, 20133 Milan, Italy. E-mail: [email protected] Department of Agricultural and Environmental Sciences – Production, Landscape, Agroenergy (DiSAA), Università degli Studi di Milano, 20133 Milan, Italy
www.soci.org
© 2014 Society of Chemical Industry
Monitoring of grape withering using visible NIR
www.soci.org
attention to optics and chemometrics as well as applications, and identified areas where more research is required. Many studies reported in the literature have evaluated the applicability of NIR spectroscopy to obtain a complete check on fruits. For example, several applications to estimate the ripeness parameters of different fruit species can be found, especially on grapes12 – 17 and leaves.18 Moreover, in recent years, interest has shifted towards the development of portable systems which can also be used in orchards and during postharvest. NIR instruments require a reliable hardware system and complex mathematical techniques to explain chemical information encoded in spectral data.19 The most commonly used chemometric techniques are (i) spectral preprocessing, to remove any irrelevant information, (ii) principal component analysis (PCA), to perform qualitative data analysis and (iii) partial least square (PLS) regression analysis, to obtain a quantitative prediction of interesting parameters.11,20 – 22 Only a few studies have been published regarding the application of non-destructive techniques for withering monitoring. An NIR acousto-optic tunable filter device (1100–2300 nm) was tested by Bellincontro et al.23 on Aleatico (Vitis vinifera L.) grapes, dehydrated at 20 ∘ C, 45% relative humidity and 1.5 m s−1 air flow in a small-scale thermo-conditioned tunnel, in order to predict total soluble solids and moisture content. The aim of the work was to test an optical portable system (visible NIR spectrophotometer) in the wavelength range 400–1000 nm for prediction of quality parameters of grape berries during withering (not forced dehydration). The chemometric analysis was focused on the correlation between the visible NIR spectra and the technological parameters most representative and interrelated with the process (firmness and soluble solids content, SSC), in order to validate the effectiveness of the system.
MATERIALS AND METHODS Sampling Sampling was performed six times on cv. Corvina (Vitis vinifera L.) berries stored in an experimental natural drying fruit room (fruttaio, barn) equipped with a monitoring system for environmental conditions, located in the Valpolicella winemaking area (Italy), from 30 September to 1 December 2012. During withering, environmental conditions change. Temperature and relative humidity (RH) start from 16 ∘ C and 65% RH and reach 9 ∘ C and 75% RH after 70 days. The optimal withering conditions were determined from previous studies which allowed us to identify the best combination of temperature and RH during the process, and in particular the best withering kinetics able to ensure high-quality withered grapes. For each sampling date 50 berries were randomly selected from bunches located in different points of the barn (barn dimensions were 500 × 600 cm). A total of 300 red berry samples (Vitis vinifera L. cv. Corvina), were analyzed. On selected berries non-destructive analysis was carried out using a portable commercial visible NIR spectrophotometer (two acquisitions were performed in the equatorial region and then averaged); physicochemical analyses were then performed.
J Sci Food Agric 2015; 95: 3144–3149
reflection measurement; (iii) a spectrophotometer; (iv) hardware for data acquisition and instrument control; (v) a battery as the power supply. Spectra were acquired in reflectance mode, without any sample preparation. Light radiation was guided from the light source to the sample through a Y-shaped, bidirectional fiber optic probe (Ocean Optics) in which seven fibers are arranged in a six-around-one configuration, each with a diameter of 600 μm. The six external fibers guide the light from the light source to the sample, while the single central fiber brings it back from the sample to the spectrophotometer. The tip of the optical probe was equipped with a soft plastic cap to ensure contact with the sample’s skin during measurements, while minimizing environmental light interference (Fig. 1). The integrated spectrophotometer was equipped with diffractive grating for spectral measurements optimized in the range of 400–1000 nm and a CCD sensor with a 2048 pixel matrix, corresponding to a nominal resolution of 0.3 nm. A calibration procedure was designed in order to minimize the effects due to possible changes in the light source or detector. The system requires the compulsory use of dark and white references (White Reflectance Standard by Edmund Optics, Barrington, NJ, USA) to finalize the calibration procedure. Furthermore, to avoid the risk of signal stability problems, the acquisition of five spectra for each measure, immediately averaged, was performed. Physical and chemical analyses The SSC was measured using a digital pocket refractometer (DBX-55 Atago, Tokyo, Japan). A few drops of the juice for each berry were placed on the refractometer sensor, previously calibrated with distilled water. The instrument gave the result directly in ∘ Brix. Firmness was measured on individual fruit using a rapid non-destructive instrument. The firmness test was performed using a Durofel® (CTIFL Copa Technologie, Saint Etienne du Grès, France) dynamometer with a bolt of 3 mm diameter (0.10 cm2 ), on a scale of 1 (soft) to 60 (firm). Results were expressed as Durofel index (DI).24 Weight loss determination was performed using an analytical balance (model PE 6000, Mettler-Toledo, Zurich, Switzerland). Data processing Mean, standard deviation (SD) and confidence intervals (𝛼 = 0.05) were calculated from the SSC, firmness and weight loss data. Chemometric analysis was performed using The Unscrambler®
© 2014 Society of Chemical Industry
wileyonlinelibrary.com/jsfa
3145
Portable visible NIR device Spectral acquisitions were performed on grape berries directly in barns using a commercial visible NIR spectrophotometer (Jaz series, Ocean Optics, Winter Par, FL, USA) operating between 400 and 1000 nm. The Jaz system consists of five components: (i) a visible NIR lighting system (halogen lamp); (ii) a fiber optic probe for
Figure 1. Detail of the fiber optic probe during spectral acquisitions on grape berry with vis/NIR portable system.
www.soci.org 9.8 software package (CAMO ASA, Oslo, Norway). Moving average smoothing (15 nm wide window), first and second derivative treatments were applied to visible NIR spectra, aimed at reducing noise before building the calibration models. Derivatives were performed using the Norris Gap transformation (11 nm wide window). PCA was performed on visible NIR spectra to examine sample grouping in order to discriminate samples and to identify outliers. PCA is a linear and unsupervised procedure that permits useful information to be extracted from the data, to explore the data structure and the relationship between objects and the global correlation of the variables.25 The visible NIR spectra were correlated with quality parameters (SSC and firmness) using the PLS regression algorithm. Models were validated using both internal validation technique (cross-validation) and test set validation with an independent validation set. Cross-validation is an internal validation method in which some samples, in this case only one (leave-one-out cross-validation) are omitted from the calibration and used for validation. This is repeated until all samples have been omitted once. To evaluate model accuracy, the coefficient of determination in calibration (R2 cal ), root mean standard error of calibration (RMSEC), coefficient of determination in validation and cross-validation (R2 val and R2 cv ), root mean standard error of validation and cross-validation (RMSEP and RMSECV) were calculated. Finally, for each elaborated model, the ratio performance deviation (RPD) was calculated. RPD is defined as the ratio between the standard deviation of the response variable and RMSE in validation. An RPD ratio of less than 1.5 indicates incorrect predictions and the model cannot be used for further prediction. An RPD between 1.5 and 2 means that the model can discriminate low from high values of the response variable; a value between 2 and 2.5 indicates that coarse quantitative predictions are possible, and a value between 2.5 and 3 or above corresponds to good and excellent prediction accuracy, respectively.26 – 28 The developed PLS models were also tested also using independent validation sets, randomly created, and comprising 40% of the total samples. Optimum calibrations were selected based on minimizing the RMSEP and the RMSECV, and maximizing the RPD.
RESULTS AND DISCUSSION The average measured raw spectra of the six sampling times during the unforced withering process are shown in Fig. 2, with
3146
Figure 2. Average raw spectra of grapes grouped by different sampling times during withering (from t0 to t5). Vertical bars indicate the standard deviation within each group at different wavelengths.
wileyonlinelibrary.com/jsfa
R Beghi et al.
bars indicating the standard deviation within each group at different wavelengths. As expected, the spectra exhibit significant differences among times, with dramatic changes in the visible NIR bands occurring from time zero (t0) to the last withering stages (t3–t5), which is especially linked to firmness modification with loss of berry water, and consequently loss of weight and turgidity, during withering (Fig. 3). In particular, this leads to a strong decrease in reflectance in the visible band associated with the anthocyanin absorption peak centered on 540 nm. Changes in spectra obviously reflect modifications in quality parameters during withering. Figure 3 also summarizes the evolution of measured parameters (SSC and firmness) during the process, compared with the weight loss trend. SSC and firmness are time dependent; in fact, correlation coefficients of R2 = 0.98 and R2 = 0.96 for SSC and firmness, respectively, were noted. The natural loss of water causes an increase of SSC, from 24.0 ± 0.8 ∘ Brix to 29.9 ± 0.5 ∘ Brix. Regarding firmness, on the contrary, a decrease from 34.8 ± 2.9 DI to 19.0 ± 0.9 D.I. was detectable. In order to evaluate the ability of the visible NIR device to differentiate samples during the withering process, data were elaborated by PCA performed on the spectral matrix. Figure 4 shows that the PCA scores plot in the plane defined by the first two principal components (PC1 and PC2) accounted for 96% of the total variance. A clear distribution of samples along PC1 and PC2 according to the sampling times during withering was found. In particular, the grape spectra of the first four sampling points (from t0 to t3, corresponding to about 40 days of withering) are distributed along the PC1. This component explains 90% of the total variance and, in fact, during the first 40 days the most evident changes to the berries occurred, in accordance with the weight loss trend (Fig. 3). Conversely, the spectra of the samples with sampling times from t3 to t5 (corresponding to the last 30 days of withering) are distributed along PC2, which explains 6% of the total variance. Over the last 30 days, in fact, a slowdown of the withering process is noted, with less evident changes in berry characteristics. Descriptive statistics and the statistics related to the PLS models obtained by visible NIR spectroscopy for analyzed indices (SSC and firmness) during withering are shown in Tables 1 and 2. Average data are shown, considering all the sampling times. Different spectral pretreatments were tested. Table 1 reports PLS models elaborated using internal validation with cross-validation technique, whereas Table 2 shows PLS models elaborated using an independent test set for validation. The models developed for SSC show a fairly good determination coefficient in calibration, ranging from 0.67 to 0.75. In cross-validation R2 ranged from 0.55 to 0.68 and the values of RPD were slightly less than 2 (1.67–1.87). The best model was developed using only the smoothing pretreatment of the spectra (Table 1). Results regarding models elaborated using an independent validation set were similar, with the same RPD values in validation. The model derived from smoothed spectra was confirmed to be the best one (Table 2). The performance of the regression models obtained for SSC prediction showed no particular differences from the results reported in the literature on other fruits.11 Bellincontro et al.23 tested an acousto-optical tunable filter (AOTF)-NIR spectrophotometer (1100–2300 nm) on Aleatico grape during controlled dehydration, obtaining good results for SSC prediction (R2 = 0.93 in cross-validation and R2 = 0.92 using an independent validation set). These results are better than those obtained in the present work. This is mainly related to the presence of the two dominant peaks, broad due to water, near 1440 and 1930 nm in the NIR region.29 The spectral range considered in the present work contain only the weaker third overtone of the OH bond (760 nm). Nevertheless, the choice to use a cheaper portable
© 2014 Society of Chemical Industry
J Sci Food Agric 2015; 95: 3144–3149
Monitoring of grape withering using visible NIR
www.soci.org
Figure 3. Evolution of SSC and firmness compared with the weight loss trend during withering. Equations of the trend lines and R2 for SSC and firmness are shown. Vertical bars indicate the confidence interval (𝛼 = 0.05) for each sampling time.
Figure 4. Scores plot of PC1 versus PC2 of PCA on grape spectra, divided by sampling times during withering process (from t0 to t5).
J Sci Food Agric 2015; 95: 3144–3149
CONCLUSIONS A portable optical device (visible NIR spectrophotometer in the wavelength range 400–1000 nm) for single-sample, non-destructive and rapid prediction of grape quality parameters directly in barns during the withering process was tested. In particular, the study was performed to estimate SSC and firmness of grape berries. The prediction capabilities of the system for both parameters were evaluated through the elaboration of prediction models based on multivariate regression techniques. Preliminary results were encouraging and the robustness of prediction models was satisfactory when tested with an independent validation set, with no significant losses of predictive power. Further studies are needed to confirm these early results, and the use of more accurate instrumentation for the reference data is desirable. Moreover, the investigation of wavelengths in order to highlight and select the most informative bands could improve the prediction capabilities of the models to monitor grapes during withering. The work has
© 2014 Society of Chemical Industry
wileyonlinelibrary.com/jsfa
3147
visible NIR spectrophotometer, compared with a conventional bench-top NIR device, is due to a potential future affordable use of the tested instrument even by small wineries. Similar results to the authors’ findings were obtained by Guidetti et al.30 using a portable visible NIR system (450–980 nm) for fresh berries of red Nebbiolo grapes, with R2 cv = 0.72 and RMSECV = 0.79 ∘ Brix for SSC. Finally, Larrain et al.31 achieved an RMSECV ranging from 1.01 to 1.27 ∘ Brix using a portable NIR device (640–1300 nm) on wine grapes under field conditions. The firmness data showed coefficients of determination in calibration from 0.51 to 0.62 and in cross-validation from 0.51 to 0.60. Fair RPD values were obtained in cross-validation, ranging from 1.64 to 1.71. The best model was developed using the first derivative pretreatment, thanks to a smaller number of latent variables (LV) with respect to the model derived from smoothed spectra (Table 1). Results were partially confirmed using an independent validation set, with R2 cal ranging from 0.64 to 0.72 and R2 val from 0.49 to 0.56. RPD values in validation were similar to those obtained in cross-validation and also in this case the best model was calculated starting from first derivative pretreatment. The analysis of firmness through the
application of visible NIR and NIR spectroscopy often shows difficulties, as highlighted by some published studies.32,33 These difficulties are mainly due to the extreme variability of this parameter among berries, particularly during a process with high variation in textural characteristics like withering. Other factors are the high instrumental error of the portable dynamometer and, also, the difficulty of calibrating a model for the estimation of an index not directly associable with a chemical species (and consequently the absorption bands of those chemical bonds). For these reasons, the predictive capabilities of the model for this parameter are usually poor. In this work, however, an R2 of about 0.6 in validation and RPD slightly less than 2 show the possibility of distinguishing between high and low firmness values.27,28 Similar results were obtained for firmness prediction on intact olives by Beghi et al.34 and by Kavdir et al.,35 using in this case FT-NIR spectroscopy in the wavelength range 780–2500. No results can be found in the literature regarding the application of visible NIR spectroscopy for the prediction of textural features of grapes during the withering process.
www.soci.org
R Beghi et al.
Table 1. Descriptive statistics and statistics of the PLS models elaborated on visible NIR spectra to estimate qualitative parameters of grape during withering, using leave-one-out cross-validation technique Calibration model Quality parameter
Mean
SD
Pretreatment
LV
n
SSC (∘ Brix)
27.8
2.7
Firmness (DI)
25.2
6.8
Smoothing Der1 Der2 Smoothing Der1 Der2
15 11 5 5 4 2
289 286 282 265 263 263
R2
Cross-validation model R2
RMSEC
0.75 0.69 0.67 0.62 0.62 0.51
1.28 1.37 1.39 3.87 3.86 4.12
RMSECV
0.68 0.62 0.55 0.60 0.60 0.51
1.46 1.54 1.63 3.97 3.97 4.15
RPD 1.87 1.77 1.67 1.71 1.71 1.64
Table 2. Descriptive statistics and statistics of the PLS models elaborated on visible NIR spectra to estimate qualitative parameters of grape during withering, using an independent test set (40% of total samples) for validation Calibration model Quality parameter
Mean
SD
Pretreatment
SSC (∘ Brix)
27.8
2.7
Firmness (DI)
25.2
6.8
Smoothing Der1 Der2 Smoothing Der1 Der2
LV 14 10 5 4 3 4
n
R2
171 173 181 146 149 155
0.79 0.67 0.69 0.64 0.66 0.72
demonstrated the applicability of visible NIR spectroscopy as a rapid technique for the analysis of grape quality during withering. This could provide the sector with simple and inexpensive optical systems which can be used to monitor the withering degree of grapes for better management of the wine production process. These instruments could be an effective solution also for small wineries seeking postharvest methods and sorting systems to improve the quality of their production.
ACKNOWLEDGEMENTS The authors would like to acknowledge Masi Agricola SpA in Gargagnago di Valpolicella (Verona, Italy) for their collaboration.
REFERENCES
3148
1 Domizio P and Lencioni L, Vin Santo. Adv Food Nutr Res 63:41–100 (2011). 2 Giordano M, Rolle L, Zeppa G and Gerbi V, Chemical and volatile composition of three sweet white passito wines. J Int Sci Vigne Vin 43:159–170 (2009). 3 Ribereau-Gayon P, Dubourdieu D, Doneche B, and Lonvaud-Funel A, Handbook of Enology: The Microbiology of Wine and Vinifications. Wiley, Chichester (2006). 4 Barbanti D, Mora B, Ferrarini R, Tornielli GB and Cipriani M, Effect of various thermo-hygrometric conditions on the withering kinetics of grapes used for the production of ’Amarone’ and ’Recioto’ wines. J Food Eng 85:350–358 (2008). 5 Muganu M, Bellincontro A, Barnaba FE, Paolocci M, Bignami C, Gambellini G et al., Influence of bunch position in the canopy on berry epicuticular wax during ripening and on weight loss during postharvest dehydration. Am J Enol Vitic 62:91–98 (2011). 6 Costantini V, Bellincontro A, De Santis D, Rotondi R and Mencarelli F, Metabolic changes of Malvasia grapes for wine production during postharvest drying. J Agric Food Chem 54:3334–3340 (2006). 7 Bellincontro A, De Santis D, Botondi R, Villa I and Mencarelli F, Different postharvest dehydration rates affect quality characteristics and volatile compounds of Malvasia, Trebbiano and Sangiovese grapes for wine production. J Sci Food Agric 84:1791–1800 (2004).
wileyonlinelibrary.com/jsfa
RMSEC 1.24 1.45 1.49 3.55 3.59 3.33
Validation model RPD 2.20 1.88 1.83 1.92 1.89 2.04
n
R2
117 117 91 103 101 98
0.62 0.59 0.52 0.56 0.56 0.49
RMSEP 1.46 1.60 1.64 3.87 3.79 4.15
RPD 1.87 1.71 1.66 1.76 1.79 1.64
8 Serratosa MP, Lopez-Toledano A, Merida J and Medina M, Changes in color and phenolic compounds during the raisining of grape cv. Pedro Ximenez. J Agric Food Chem 56:2810–2816 (2008). 9 Rolle L, Torchio F, Giocosa S and Gerbi V, Modifications of mechanical characteristics and phenolic composition in berry skins and seeds of Mondeuse wine grapes throughout the on-vine drying process. J Sci Food Agric 89:1973–1980 (2009). 10 Wang W and Paliwal J, Near-infrared spectroscopy and imaging in food quality and safety. Sens Instrum Food Qual 1:193–207 (2007). 11 Nicolaï BM, Beullens K, Bobelyn E, Peirs A, Saeys W, Theron KI et al., Non-destructive measurement of fruit and vegetable quality by means of NIR spectroscopy: a review. Postharvest Biol Technol 46:99–118 (2007). 12 Lafontaine M, Freund M, Vieth KU and Negara C, Automatic fruit sorting by non-destructive determination of quality parameters using visible/near infrared to improve wine quality. I. Red wine production. NIR News 24:6–8 (2013). 13 Bellincontro A, Nicoletti I, Valentini M, Tomas A, De Santis D, Corradini D et al., Integration of nondestructive techniques with destructive analyses to study postharvest water stress of wine grapes. Am J Enol Vitic 60:57–65 (2009). 14 Fernández-Novales J, López MI, Sánchez MT, Morales J and González-Caballero V, Shortwave-near infrared spectroscopy for determination of reducing sugar content during grape ripening, winemaking, and aging of white and red wines. Food Res Int 42:285–291 (2009). 15 Cozzolino D, Dambergs RG, Janik L, Cynkar WU and Gishen M, Analysis of grapes and wine by near infrared spectroscopy. J Near Infrared Spectrosc 14:279–289 (2006). 16 Gishen M, Dambergs RG and Cozzolino D, Grape and wine analysis – enhancing the power of spectroscopy with chemometrics: a review of some applications in the Australian wine industry. Aust J Grape Wine Res 11:296–305 (2005). 17 Cozzolino D, Esler MB, Dambergs RG, Cynkar WU, Boehm DR, Francis IL et al., Prediction of colour and pH in grapes using a diode array spectrophotometer (400–1100 nm). J Near Infrared Spectrosc 12:105–111 (2004). 18 De Bei R, Cozzolino D, Sullivan W, Cynkar W, Fuentes S, Dambergs R et al., Non-destructive measurement of grapevine water potential using near infrared spectroscopy. Aust J Grape Wine Res 17:62–71 (2011).
© 2014 Society of Chemical Industry
J Sci Food Agric 2015; 95: 3144–3149
Monitoring of grape withering using visible NIR
www.soci.org
19 Cogdill RP and Anderson CA, Efficient spectroscopic calibration using net analyte signal and pure component projection methods. J Near Infrared Spectrosc 13:119–132 (2005). 20 Wold S, Sjöström M and Eriksson L, PLS-regression: a basic tool of chemometrics. Chemom Intell Lab Syst 58:109–130 (2001). 21 Naes T, Isaksson T, Fearn T and Davies T, A User-Friendly Guide to Multivariate Calibration and Classification. NIR Publications, Chichester (2002). 22 Cen H and He Y, Theory and application of near infrared reflectance spectroscopy in determination of food quality. Trends Food Sci Technol 18:72–83 (2007). 23 Bellincontro A, Cozzolino D and Mencarelli F, Application of NIR-AOTF spectroscopy to monitor Aleatico grape dehydration for Passito wine production. Am J Enol Vitic 62:256–260 (2011). 24 Chiabrando V, Giacalone G and Rolle L, Mechanical behaviour and quality traits of highbush blueberry during postharvest storage. J Sci Food Agric 89:989–992 (2009). 25 Beebe KR, Pell RJ and Seasholtz MB, Chemometrics: A Practical Guide. Wiley, New York (1998). 26 Williams PC, Variables affecting near-infrared reflectance spectroscopic analysis, in Near-Infrared Technology in the Agricultural and Food Industries, ed. by Williams PC and Norris KH. American Association of Cereal Chemists, St Paul, MN, pp. 143–167 (1987). 27 Williams PC, Implementation of near-infrared technology, in Near-Infrared Technology in the Agricultural and Food Industries,
28 29 30 31 32
33
34 35
ed. by Williams P and Norris KH. American Association of Cereal Chemists, St Paul, MN, pp. 145–169 (2001). Fearn T, Assessing calibrations: SEP, RPD, RER and R2 . NIR News 13:12–14 (2002). Büning-Pfaue H, Analysis of water in food by near infrared spectroscopy. Food Chem 82:107–115 (2003). Guidetti R, Beghi R and Bodria L, Evaluation of grape quality parameters by a simple Vis/NIR system. Trans ASABE 53:477–484 (2010). Larrain M, Guesalaga AR and Agosin E, A multipurpose portable instrument for determining ripeness in wine grapes using NIR spectroscopy. IEEE Trans Instrum Meas 57:294–302 (2008). Nicolaï BM, Verlinden BE, Desmet M, Saevels S, Theron K, Cubeddu R et al., Time-resolved and continuous wave NIR reflectance spectroscopy to predict soluble solids content and firmness of pear. Postharv Biol Technol 47:68–74 (2008). Zude M, Herold B, Roger JM, Bellon-Maurel V and Landahl S, Non-destructive tests on the prediction of apple fruit flesh firmness and soluble solids content on tree and in shelf life. J Food Eng 77:254–260 (2006). Beghi R, Giovenzana V, Civelli R, Cini E and Guidetti R, Characterization of olive fruit for the milling process by using visible/near infrared spectroscopy. J Agric Eng 44:56–61 (2013). Kavdir I, Buyukcan MB, Lu R, Kocabiyik H and Seker M, Prediction of olive quality using FT-NIR spectroscopy in reflectance and transmittance modes. Biosyst Eng 103:304–312 (2009).
3149
J Sci Food Agric 2015; 95: 3144–3149
© 2014 Society of Chemical Industry
wileyonlinelibrary.com/jsfa
