Pear quality characteristics by Vis / NIR spectroscopy

Recently, non-destructive techniques such as the Vis / NIR spectroscopy have been used to evaluate the characteristics of maturation and quality of pears. The study aims to validate the readings by the Vis / NIR spectroscopy as a non-destructive way to assess the qualitative characteristics of pear cultivars ‘Williams’, ‘Packams’ and ‘Carrick’, produced according to Brazilian conditions. The experiment was conducted at the Pelotas Federal University, UFPel, in Pelotas / RS, and the instrument used to measure the fruit quality in a non-destructive way was the NIRCase spectrophotometer (SACMI, Imola, Italy). To determine pears ́ soluble solids (SS) and pulp firmness (PF), it was established calibration equations for each variety studied, done from the evaluations obtained by a non-destructive method (NIR-Case) and a destructive method. Further on, it was tested the performance of these readings by linear regressions. The results were significant for the soluble solids parameter obtained by the Vis / NIR spectroscopy; however, it did not achieve satisfactory results for the pear pulp firmness of these cultivars. It is concluded that the Vis / NIR spectroscopy, using linear regression, allows providing reliable estimates of pears ́ quality levels, especially for soluble solids.


INTRODUCTION
The decision of the consumer to purchase a fruit is usually based on the color and the ripening stages that are related to its sweetness and firmness.The content of soluble solids (SS) and firmness (PF) are important characteristics that determine, in a large scale, the perception of pear quality by the consumer.However, the determination of these characteristics is still performed in a destructive manner (Liu et al. 2008).This method of analysis requires the destruction of the fruits and a limited number of Saranwong et al. (2004), Han et al. (2006), Fu et al. (2007), Shao et al. (2007), Nicolai et al. (2007), Harker et al. (2008), Sinelli et al. (2008), Fan et al. (2009), Pérez-Marín et al. (2009), Xie et al. (2009), Valente et al. (2009), Cynkar et al. (2009), Camps andChristen (2009) andkavdir et al. (2009).
The Vis / NIR spectroscopy has as an advantage, the possibility of determining, simultaneously, more than one quality characteristic, increasing the number of samples measurements and repeating the analysis of the same samples, by monitoring physiological changes in the fruit during the ripening process (Nicolai et al. 2007, Gabioud et al. 2008).
The spectroscopic data on the fruit internal quality are obtained by evaluating the absorption of Vis/ NIR light by functional groups at a specific wavelength (Banwell 1983).The procedure is to obtain a series of calibration samples and to evaluate the spectral data, as well as the quality characteristics of interest, which is determined by a reference method (usually destructive) (Peirs et al. 2003).
The Vis / NIR spectroscopy with wavelengths between 750 and 2500 nm has been used as a nondestructive method to determine the properties of organics products (Huang et al. 2007), and it has been successfully implemented for various fruits such as apples (Liu et al. 2007), peaches (Ying et al. 2005), kiwifruit (McGlone et al. 2002a) and mangoes (Schmilovitch et al. 2000).According to Liu et al. (2008), no information was found about models of validation for the Vis / NIR spectroscopy for soluble solids and firmness of intact pears.Furthermore, the wavelengths of 750-2500 nm are easily transmitted through the pear fruit; however, it is possible that other wavelengths successfully allow extracting information.In the present study, the Vis / NIR spectroscopy with a wavelength of 600-1000nm was used.
Therefore, the study aims to validate the readings by the Vis / NIR spectroscopy, as a nondestructive way to evaluate the pear quality of the cultivars 'Williams', 'Packams' and 'Carrick'.

MATERIALS AND METHODS
The experiment was conducted at the Laboratory of Fruticulture of Pelotas Federal University (UFPel) in Pelotas / RS, in February 2010 with 'Williams', 'Carrick' and 'Packams' pears from the Orchard of Palma Agricultural Center, at UFPel.The fruits were harvested at different stages of ripeness and from different locations, and stored in a cold chamber (0°C and 90% RH) for 30 days (each sample of the respective cultivars consisted of 40 fruits) for both the destructive evaluation by conventional methods, as well as the nondestructive evaluation to obtain the calibrations of different cultivars and to validate them.
The instrument used to determine the quality of fruits in a non-destructive way was the NIR-Case spectrophotometer (SACMI, Imola, Italy).This equipment performs the reading by transmittance.The light source consists of eight halogen lamp, and the wavelength varies between 600 and 1000nm.The readings were performed by the equipment considering the equatorial region of the fruit on opposite sides (A and B) of each pear for each cultivar determining the content of soluble solids (SS) and pulp firmness (FP) of the fruits.
To determinate the pears' SS and FP, first, calibration equations were settled for each variety studied from the evaluations by the non-destructive (NIR-Case) and destructive (reference method) methods.After obtaining the spectra from the existing equations for pears for the NIR-Case (SACM), evaluations of reference (destructive) were carried out separately considering the same sides of the fruits; evaluating the pulp firmness (PF) with a digital penetrometer with a tip diameter of 8mm (model 53 205, TR, Forli, Italy), removing the peel from both opposite sides of each fruit in the equatorial region, which the results were expressed in Newton (N); and it was also evaluated the solids soluble content (SS) with a digital refractometer (Atago PR32) in the same spots where the fruit firmness was analyzed and the PEAR QUALITY BY VIS / NIR SPECTROSCOPY results expressed in °Brix.The spectra generated by NIR-Case and the values obtained by the method of reference of the samples were calculated using the NCS (NIR CALIBRATION SOFTWARE) Version 3.0 R C 1, resulting in new calibration equations for each cultivar.The correlations obtained by the spectra with the reference values of the quality characteristics for calibrations are performed by the software that comes with the equipment.
After obtaining the calibration equations for 'Williams', 'Packams' and 'Carrick' pears, the performance was tested (the validation).It was proceeded the validation of each equation for each cultivar obtained by the SACMI software for the NIR-Case spectrophotometer.It was used linear regressions to the values obtained through the NIR-Case and the reference method for the new samples.For this effect, it was utilized the Past program version 1.82b.The statistical parameters correlation coefficient (R), mean square error of predicted values (RMSEP) and the standard error of predicted values (SEP) were used to determine the accuracy of the calibration equations.
After the validation of the calibration equations for pears, experiments were designed consisting of pears cultivars 'Williams', 'Packams' and 'Carrick', on different rootstocks to determine the characteristics of quality using the best calibration equations obtained by the NIR-Case spectrophotometer.Spectral data were obtained through the NIR-Case to determine the characteristics of the quality of 'Williams', grafted on the Champion, Meliforme and Pyrus calleryana rootstocks, 'Packams' on the D'Angers, Smyrna, Alongado, Adam's and P. calleryana rootstocks, and 'Carrick' grafted on Portugal, MC, D'Vranje, BA 29 and Inta 267.

RESULTS AND DISCUSSIONS
The calibrations from the 'Williams', 'Packams' and 'Carrick' pears were obtained as a way of estimating accurate values of soluble solids (SS) and pulp firmness (PF), with the results being close to those obtained by the reference method.Table I presents the results of statistical parameters related to the calibrations, which were subjected to further validation in order to determine the respective quality characteristics of the pear cultivars according to the Brazilian conditions (Figs. 1, 2 and 3)

Soluble solids
The characterization of the values for soluble solids (SS) of pears that are used to validate the calibration of the cultivars evaluated and obtained by the SACMI software are presented in Table II.
It is verified that the SS obtained for the validation of the calibration presented a variation between 9.60 and 12.90 ° Brix, 10.50 and 15.00 ° Brix and 10.80 and 16.20 ° Brix for 'Williams', 'Packams' and 'Carrick', respectively.The largest gap between the minimum and maximum values of SS was obtained for 'Carrick' pears, which coincides with the ranges obtained by Nicolai et al. (2008) (10.4 and 15.4 ° Brix) for 'Conference' pears and by Liu et al. (2008) for 'Fengshi' pears (SS between 8.4 and 14.1 ° Brix).When comparing the values of minimum, maximum and NICÁCIA P. MACHADO et al. the average of both methods to determine SS (NIR-Case and reference), the values are relatively close.To 'Packams' cultivar, it was observed a difference of 0.93 for the average values of SS obtained from both methods (NIR-Case and reference), being higher when compared with the evaluated cultivars, mainly with the cultivar 'Carrick' that had the smallest difference (0.08).However, there seems to be a good correlation between the two methods for determining SS of the 'Packams' pears.
The results of the linear regressions for the values of the NIR-Case and of the soluble solids measured destructively (reference method) for the validation of the calibrations are shown in Figures 1, 2 and 3.It is observed that the coefficients allow to estimate the existence of a high correlation between the two methods for SS for 'Carrick' and 'Packams' pears (R = 0.82 and 0.80, respectively).These values were higher than those observed by Nicolai et al. (2008) who obtained an R of 0.77 for 'Conference' pears using the Vis/ NIR spectroscopy with a wavelength of 780-1700nm, and by Costa et al. (2002) who obtained an R = 0.71 for 'Abbé Fetel', 'Pass Crassane', 'Conference' and 'Comice' pears .However, these values were lower than the ones reported by Liu et al. (2008) who obtained a higher correlation coefficient (R = 0.91) for 'Fengshui' pears using a wavelength of 350-1800nm, and by Sun et al. (2009) who evaluated the use of Vis / NIR with a wavelength of 345-1040nm to determine SS for 'Cuiguan' pears, obtaining an R of 0.96 and an RMSEP of 0.53.
To validate the calibration equation for 'Williams', it was obtained an R of 0.60.Although this correlation coefficient between values obtained by both methods is lower (R = 0.60) than the coefficients obtained for 'Packams' and 'Carrick', and the R values for the respective cultivars evaluated in this study were within the obtained values by other authors.Tsai et al. (2007) and Ying and Liu (2008) worked with Japanese and Chinese pears and obtained correlation coefficients for SS between 0.60 and 0.85, and SEP between 0.40 and 0.62.
For the statistical parameters correlation coefficient (R), mean square error of predicted values (RMSEP) and standard error of predicted values (SEP), which were used to determine the accuracy of the calibration equations, it was observed the lowest value of RMSEP (0.34) for 'Carrick' and the highest value of RMSEP (4.17) for the cultivar 'Packams'.The SEP values obtained for the cultivars 'Williams', 'Packams' and 'Carrick' were 1.75, 4.06 and 0.33, respectively (Table III).The value of RMSEP (0.34) obtained in the validation of the calibration of the SS to 'Carrick' pears was less than the values obtained by Nicolai et al. (2008) who observed an RMSEP of 0.44 for 'Conference' pears with a wavelength between 780 and 1700nm, and by Liu et al. (2008) who obtained 0.66 for 'Fengshui' pears with a wavelength between 350 and 1800nm.Regarding the SEP values, Costa et al. ( 2002) had a SEP of 1.15 for the SS of pears, and Tsai et al. (2007) and Ying and Liu (2008), who worked with Japanese and Chinese pears, obtained a SEP between 0.40 and 0.62.
Regarding the RMSEP and SEP, ratings of 4.17 and 4.0, respectively, were observed for the cultivar 'Packams', in spite of a good correlation coefficient (R = 0.80), which could be due to the fruit sampling for the analysis of pears of the respective cultivar being more homogeneous.Delwiche and Norris (1993) concluded that incorporating heterogeneous samples in the calibration equation improves the classification accuracy of this equation, but it decreases the accuracy of the prediction of nonheterogeneous samples.Several authors, among them, Xiaobo et al. (2007), also studied the influence of regression models used and observed differences in the values of SEP and RMSEP for the SS.
Correlation coefficients relatively high and low RMSEP values indicate that non-destructive determinations by Vis / NIR can provide good estimates for pear quality parameters, especially for soluble solids (Liu et al. 2008).The correlation coefficients of the validation of the calibrations for the cultivars studied for SS coincide or are lower or 857 PEAR QUALITY BY VIS / NIR SPECTROSCOPY higher than those obtained for other fruits.Ventura et al. (1998), using NIR (811-999nm) for SS validation, obtained an R of 0.73 and 0.75 and a SEP of 1.05 for 'Golden Delicious' apple, and an R of 0.68 and 0.71 and a SEP from 1.06 to 1.11 for 'Jonagold' apples.McGlone et al. (2002b), using 500-1100nm, obtained an R of 0.79 and a SEP of 0.5 for Royal Gala apples.Liu et al. (2007) obtained an R of 0.84 and a SEP of 0.77 for 'Fuji apples'.For 'Sweet Lady' nectarines, using 1600-2400nm and 400-1700nm, the correlation coefficients were between 0.87 and 0.93 for SS (Pérez-Marín et al. 2009).Golic and Walsh (2006) obtained correlations of 0.88 for SS of pit fruits (peaches, nectarines and plums).
To validate the calibrations for the determination of the soluble solids content of pear cultivars 'Williams', 'Packams' and 'Carrick', it was obtained satisfactory correlation coefficients, being possible to accurately estimate the quality characteristic using the Vis / NIR spectroscopy with a wavelength of 600 -1000nm.McGlone and kawano (1998) state that the wavelength between 800 and 1100nm is a good interval to estimate the content of soluble solids.
In general, it is suggested that the estimation of the SS by the NIR-Case is enhanced with new data, i.e., new calibration equations, in order to obtain a better precision in the estimation of the data.It is also suggested to increase the number of samples to be used for calibration with a larger range of values for the characteristic to be studied in order to increase the heterogeneity of the sample, and the greater the range, the better the consistency of the calibration.In the literature, the number of samples is quite variable, and it can be seen samples of 40 fruits (Gabioud et al. 2008), 80 fruits (Liu et al. 2008), 200 fruits (Costa et al. 2002), 340 fruits (Ventura et al. 1998) and 420 fruits (Nicolai et al. 2008).

Pulp firmness
For the values of the pear pulp firmness (PF), there was a range between 26.38 and 55.90 N for 'Williams', 31.48 and 71.39 N for 'Packams', and 35.01 and 62.47 N for 'Carrick'.The largest gap between the minimum and maximum values was observed for 'Packams', being this interval smaller than the range obtained by Nicolai et al. (2008) for 'Conference', with a range between 9.9 and 110.7 N for pulp firmness and considerably larger than the range observed by Liu et al. (2008) for 'Fengshi' (2.9 to 13.7 N).When comparing the average values for the firmness obtained by the reference method and the NIR-Case, it is observed that the values of the NIR-Case were lower, except for the cultivar 'Carrick', the others were smaller than those obtained by the reference method.The smallest difference between the mean values for the pulp firmness obtained by the two methods was observed for the cultivar 'Carrick' (difference of 0.12), and the highest difference was observed for 'Packams' (0.90) (Table II).
With respect to the linear regression between the values of firmness determined by the two methods for 'Williams', 'Packams' and 'Carrick' pears, correlation coefficients of R = 0.35, R = 0.26 and R = 0.12 were obtained, respectively (Figs. 1, 2 and 3).The obtained values of R are lower than the results observed by Nicolai et al. (2008) in 'Conference' pears with an R= 0.59, and the same authors consider these values of R unsatisfactory for determining the firmness by spectroscopic techniques.However, these results are very different from those obtained by Costa et al. (2002) who found an R = 0.71 in four pear cultivars (Abbé Fetel, Crassane Pass, Conference and Comice), and Liu et al. (2008), who obtained an R = 0.85 for 'Fengshui' pears with an interval of wavelength between 350 and 1800nm, found greater values that have been obtained in this study.For other fruits, authors such as Valero et al. (2004) obtained an R = 0.77 for kiwis.Gabioud et al. (2008), using NIR-Case (SACM), observed a correlation coefficient of 0.76 for apples firmness.Gómez et al. (2006), using a wavelength of 400-2350nm, observed a correlation of 0.69 to determine tangerines firmness.NICÁCIA P. MACHADO et al.Table III shows the statistical parameters concerning the validation of the calibrations for the determination of pear firmness.In which, it was obtained the lowest RMSEP for the 'Carrick' cultivar (RMSEP = 5.29 N).This value was lower than that observed by Nicolai et al. ( 2008) (RMSEP = 10.4N) for pears, by 'McGlone' and 'kawano' (1998) for kiwis (RMSEP = 7.0 N) and by Gómez et al. (2006) for tangerines (RMSEP = 8.53 N), and higher RMSEP = 1.23 N obtained by Liu et al. (2008).For 'Packmas' cultivars, an RMSEP = 34.48N was obtained, which is a value considered relatively high for the determination of firmness.According to Peirs et al. (2002), the prediction standard errors (RMSEP and SEP) ranged between 8 and 12N; however, Lu and Peng (2006) obtained SEP values between 15 and 20N for peaches firmness using a wavelength between 500-1000nm.The value of SEP was also the lowest value in calibrations for pears firmness of 'Carrick' (SEP = 5.1 N), being this value lower than that obtained by Costa et al. (2002) (SEP of 8.82 N) for pears and by Lu et al. (2000) (SEP of 6.2 N) for 'Golden Delicious' apples and by Schmilovitch et al. (2000) for mangoes with SEP of 17.14 N.
For Liu et al. (2008), correlation coefficients relatively high and RMSEP and SEP with values considered low indicate that nondestructive determinations by Vis / NIR can provide good estimates for pear quality parameters.In this study, although the statistical parameters (RMSEP and SEP) for the validation of calibrations for the firmness of the respective cultivars are within the limit when compared to those obtained by other authors, the correlation coefficients (R) obtained showed low values.The cause to justify the low values of RMSEP and SEP (indicating a good estimate) even with low correlations may be due to the use of heterogeneous samples to obtain the calibrations and to validate them, which confirms the wide range of values obtained for the firmness of the fruits of different cultivars evaluated.
According to Delwiche and Norris (1993), heterogeneous samples in the calibration equations improve the classification accuracy of this equation, but they decrease the accuracy of prediction of nonheterogeneous samples.
The weak correlations obtained for the pulp firmness of 'Williams', 'Packams' and 'Carrick' cultivars may be due to several factors such as the range of the wavelength used by the NIR-Case (between 600 and 1000nm) not being the most suitable for extracting information on the spectrum bands that are related to the pear firmness of these cultivars, as verified by Swierenga et al. (2000).Moreover, according to Harker et al. (1996), in the reference method used to determine the firmness by a penetrometer, the readings may vary according to the evaluator; in other words, the evaluator may influence the measurements performed by the equipment.Nicolai et al. (2006) suggests the use of non-destructive methods of impact or vibration to determine this trait.
According to zerbini (2006), the Vis / NIR spectroscopy has been used to estimate quality characteristics such as soluble solids, firmness, dry matter and acidity.In general, the estimation of SS is satisfactory and consistent, while for the remaining properties, the results are quite variable.Nicolai et al. (2008) argues that the Vis / NIR spectroscopic techniques are not valid to estimate the firmness of pears.
In the regression methods used for data processing, two or more instrumental responses are related to the property of interest.These methods allow the analysis even with some interference, since they are present in the calibration samples (Brereton 2000).In this study, the linear regression method used for processing data for the validation of the calibration of pears cultivars provided significant results for the determination of SS by the Vis / NIR spectroscopy; however, no good results were obtained for pear firmness of these cultivars.Based on the results obtained in this study, to validate the calibration for determining the pear quality characteristics by the Vis / NIR spectroscopy to be satisfactory only for soluble solids, they were also tried in 'Williams', 'Packams' and 'Carrick' pears, grafted on different rootstocks using the best calibration.
Table IV presents the results of the descriptive statistics of the values of soluble solids content of 'Williams', 'Packams' and 'Carrick' pears on different rootstocks determined by the Vis / NIR spectroscopy using calibrations obtained by NIR-Case.And, as observed, the variation between the minimum and maximum values of SS for each cultivar grafted on different rootstocks has values close to those obtained by the reference method for the validation of the calibrations used for the determination of this quality characteristic.It can be observed that the averages for each cultivar are close to the averages in the validation.Thus, it is confirmed that the use of them determines the evolution of the ripe fruit and the most appropriate harvest time.

CONCLUSIONS
It can be concluded that it is possible to use a nondestructive method for the Vis / NIR spectroscopy to determine pear quality characteristics.With the use of linear regression, relations were established between spectral parameters obtained by the Vis / NIR spectroscopy and internal fruit quality, which make possible to provide reliable estimates of the quality indexes for pears, especially for soluble solids.Palavras-chave: calibração, firmeza, pereira, sólidos solúveis, validação.

Figure 1 .
Figure 1.Relation between the values of soluble solids and pulp firmness estimated by NIR-Case and the values obtained for each characteristic by the destructive method (of reference) used to validate the calibration for 'Williams' pears.UFPel, 2010.

Figure 3 .
Figure 3. Relation between the values of soluble solids and pulp firmness estimated by NIR-Case and the values obtained for each characteristic by the destructive method (of reference) used to validate the calibration for 'Carrick' pears.UFPel, 2010.

Figure 2 .
Figure 2. Relation between the values of soluble solids and pulp firmness estimated by NIR-Case and the values obtained for each characteristic by the destructive method (of reference) used to validate the calibration for 'Packams' pears.UFPel, 2010.

TABLE I Statistical parameters related to the calibrations for soluble solids (SS) and pulp firmness (PF) of pears cultivars 'Williams', 'Packams' and 'Carrick' obtained with the NIR-Case. UFPel, 2010.
VALIDATION OF THE CALIBRATIONS TO THE DETERMINATION OF SOLUBLE SOLIDS AND PULP FIRMNESS OF PEARS PEAR QUALITY BY VIS / NIR SPECTROSCOPY