QUALITY EVALUATION OF “TARDIVO DI CIACULLI” MANDARINS IN POST-HARVEST PROCESSING ON AN INDUSTRIAL SCALE USING A PORTABLE VIS/NIR DEVICE

(1)

Q UALITY E VALUATION OF “T ARDIVO DI C IACULLI ” M ANDARINS IN P OST -H ARVEST P ROCESSING ON AN

I NDUSTRIAL S CALE U SING A P ORTABLE V IS /NIR D EVICE

M. Vallone, F. Bono, M. Alleri, E. Barone, P. Catania

ABSTRACT. Vis/NIR technology is widely used today to quickly and non-destructively evaluate fruit and vegetable quali- ties, and many applications have been found since the 1990s. However, no industrial-scale applications can further con- solidate the use of non-destructive techniques in post-harvest processing. This study aims to test the possibility of applying vis/NIR technology in a modern citrus-processing plant to assess the damage that the fruits eventually suffer when they are processed on an industrial scale and the evolution of their key quality parameters in a period of 10 days after harvest.

The spectral acquisitions were performed using a portable vis/NIR device, which operated in the wavelength range of 600 to 1000 nm. The firmness, pH, and soluble solids content (SSC) were studied for “Tardivo di Ciaculli” mandarin on a total of 1800 fruits. The results show that the vis/NIR device can predict 96% of the total variability of the observed values for fruit firmness; however, the insignificance of the coefficients corresponding to the different sampling points of the pro- cessing plant, in the early stages of time, denotes the poor ability of the device to properly detect firmness in the time points closer to the fruit processing. The vis/NIR device explains 93% of the variability of the observed pH and SSC val- ues.

Keywords. Mandarin, Post-harvest, Processing plant, Vis/NIR.

ruits and vegetables must maintain high organolep- tic and nutritional characteristics, but some phases of their processing can cause mechanical damage to the product. Moreover, this type of damage occurs with a certain time lag and appears in an advanced stage of marketing or on the consumer table. The introduction of non- invasive instruments to assess fruit quality can help to solve the problem. Currently, identifying any critical points in the early stages of post-harvest processing and prolonging the marketing of fresh product while improving the image of authenticity and freshness are targets of primary importance.

Considering the growing interest from the market, extensive research has been conducted on the preservation of postharvest quality in fruits and vegetables (Ortiz and Tor- regrosa, 2014; Giovanelli et al., 2014).

The selection of harvested fruits mainly concerns the color, size, and presence of defects; furthermore, product characteristics such as soluble solids content (SSC), firm-

ness, and acidity are usually determined using destructive methods. However, near-infrared spectroscopy (vis/NIR) has been developed and used as a non-destructive technique to quantitatively and qualitatively characterize many types of fruit. This technique has begun a new approach of market segmentation to evaluate both fresh and processed fruits (Nicolai et al., 2007) and can classify and certify products based on the quality criteria before harvest and during and after packaging.

Several researchers evaluated the starch index, SSC, water content, acidity, texture, and other physiological properties using near-infrared spectroscopy on citrus and orange (Steuer et al., 2001; Miller and Zude-Sasse, 2004; Cayuela, 2008; Lu et al., 2008; Zude et al., 2008; Cayuela and Weiland, 2010; Liu et al., 2010a ; Zheng et al., 2010;

Magwaza et al., 2012; Sanchez et al., 2013), mandarin (Kawano et al., 1993; Fraser et al., 2003; McGlone et al., 2003; Guthrie et al., 2005 a, 2005b; Gomez et al., 2006;

Sun et al., 2009; Liu et al., 2010b; Magwaza et al., 2014), tomato (Slaughter et al., 1996), mango (Saranwong et al., 2004), kiwifruit (Osborne et al., 1999), and apple (Lammer- tyn et al., 1998; Park et al., 2003; Menesatti et al., 2009;

Beghi et al., 2012, 2014) with the advantage of fast and non-destructive evaluation of the product quality. Most of these studies concern laboratory applications of vis/NIR technology.

This study aims to test the potential use of vis/NIR technology in a modern citrus-processing plant to assess the damage that the fruits eventually suffer because of industrial-scale processing and the evolution of their key quality parameters in a period of 10 days after harvest.

Submitted for review in April 2015 as manuscript number MS 11328;

approved for publication by the Machinery Systems Community of ASABE in December 2015.

The authors are Mariangela Vallone, Research Associate, Department of Agricultural and Forest Sciences, Filippa Bono, Research Associate, Department of Economic Business and Statistical Sciences, Maria Alleri, Doctoral Student, Department of Agricultural and Forest Sciences, Ettore Barone, Full Professor, Department of Agricultural and Forest Sciences, and Pietro Catania, Associate Professor, Department of Agricultural and Forest Sciences, University of Palermo, Palermo, Italy. Corresponding author: Mariangela Vallone, Department of Agricultural and Forest Sciences, University of Palermo, Viale delle Scienze ed. 4, 90128 Palermo, Italy; phone: 0039-9123865609; e-mail: mariangela.vallone@

unipa.it.

F

(2)

M

ATERIALS AND

M

ETHOD

The research was performed in 2014 in mandarin or- chards in the district of Palermo, Italy, that were included in the Tardivo di Ciaculli mandarin consortium. The “Tar- divo di Ciaculli” mandarin belongs to the family Rutaceae, species Citrus reticulata. It comes from an Avana mandarin spontaneous bud mutation and is a late-ripening variety (January to March).

TARDIVO DI CIACULLI MANDARIN

PROCESSING PLANT

The Tardivo di Ciaculli mandarin processing plant pro- cesses the fruits supplied by the producers of the consortium. The manually harvested mandarins arrive at the consortium in plastic boxes of 18 to 20 kg capacity. After weighting, the boxes are transported to the processing plant, which has a work capacity of 4 t h^-1 (fig. 1).

Four sampling points, which are called “tests” in the present work (table 1), were used in the processing plant where the fruits could be subject to shock and damage. The first point (A) was located immediately after harvest, when the fruits are placed in plastic boxes for transport to the processing plant. The second point (B) was located immediately after fruit unloading in the processing line and spe- cifically after box overturning, when the fruits fall from a height of approximately 0.30 m. The third point (C) was located immediately after the roller bench for manual selection before the conveyor belt, which carries the fruits to the electronic sizer. At this point, the fall height of the fruits is approximately 0.20 m because of the difference in height between the roller bench and the conveyor belt. The fourth

point (D) was located after sizing, when the fruits are di- rected to the output tapes and placed in cardboard baskets for transport to market.

PORTABLE VIS/NIRDEVICE

The spectral acquisitions were performed using a vis/NIR system (model NCS001, Sacmi, Imola, Italy) that was operated in the wavelength range of 600 to 1000 nm (table 2).

The system consists of five elements: a lighting system, a power pack for lighting, a spectrophotometer, a PC for the user interface, which is responsible for data display and management parameters, and a fan to control the temperature. A control keyboard and a monitor complete the system

Figure 1. Operation of the Tardivo di Ciaculli mandarin processing plant. Tests A, B, C, and D are described in table 1.

Table 1. Sampling points along the mandarin processing plant.

Test Description

A After harvest

B Fruit unloading in the processing plant

C After manual selection

D After electronic sizer

Table 2. Technical features of the NCS001 vis/NIR system.

Parameter Description Measurement method Transmittance

Wavelength range 600 to 1000 nm Light source Halogen lamps

Display Color TFT with touch screen Measuring time 6 ms to 2 s

External communication USB cable, Ethernet 10/100 Mb cable, wireless connection

Ambient conditions for use Temperature = 10°C to 35°C; relative humidity = 25% to 80%

Power supply 100 to 240 V, 50/60 Hz 2.8 to 1.2 A Dimensions 400 × 300 × 200 mm

Weight 8.50 kg

(3)

(fig. 2). The measurements were acquired using dedicated software (NCS software package, Sacmi, Imola, Italy).

The system is based on the projection of an intense light beam with a near-infrared frequency band through the body of the product. The resultant light is collimated into a single narrow beam and analyzed to determine the parameters of interest. Two acquisitions were performed for each fruit along the equator region on opposite sides. A 100 mm² Teflon disk was used as the optical reference standard for the system, since Teflon has low reflectance and its light- scattering characteristics are similar to those of the samples.

DESTRUCTIVE PHYSICAL AND CHEMICAL ANALYSES

Fruit firmness was determined using a mechanical dy- namometer (DPS 5R, Imada, Northbrook, Ill.), which was connected to an electronic stand (MX2-500N-L, Imada), and a PC for downloading data. Fruit compression was obtained using a cylindrical steel plate that was 50 mm in diameter, the surface of which was disposed orthogonally on the minor axis of the mandarin. The test speed was maintained at 0.167 mm s^-1 during the tests (Catania et al., 2014, 2015). The compression force was continuously recorded during the entire process; the peak force (N) was the maximum registered force during compression (Beghi et al., 2014). The measurements were performed on each entire fruit without skin.

The pH value was measured using a portable pH meter (MM40 multimeter, Crison Instruments, Barcelona, Spain).

The SSC was measured using a portable refractometer (MR32ATC, Milwaukee Instruments, Rocky Mount, N.C.) after the spectrophotometer acquisitions of each fruit and by directly squeezing the juice onto the refractometer, which was previously calibrated with distilled water. The results are expressed in °Brix.

S

AMPLING

SAMPLING FOR VIS/NIRDEVICE CALIBRATION

The calibration procedure for the vis/NIR device was

performed with reference to the parameters that we studied:

firmness (N), pH, and SSC (°Brix). Thus, the fruits were sampled at three different stages: in the field immediately after manual harvest, in the input of the processing plant, and in the output of the processing plant. In this way, a sample of fruits that would be damaged during the entire process from field to packing was considered.

For the calibration procedure, a total of 1800 fruits were analyzed on the day of harvest and at five successive time points: 48, 96, 144, 192, and 240 h after harvest (100 fruits per time point). The first time point (48 h) corresponds to the period in which the mandarins reach the major domestic and foreign markets. The other time points cover the period in which the fruit is consumed (10 days). During the analysis, the fruits were stored in the laboratory at 10°C and 70%

relative humidity.

SAMPLING FOR THE PROCESSING PLANT STUDY

After the calibration phase, a sample of 25 fruits was taken at each sampling point (described in table 1) for a total of 100 fruits. They were subjected to spectrum acqui- sition using the vis/NIR device, which was previously calibrated, and subsequently to laboratory analyses to destructively determine the firmness, pH, and SSC. The analyses were performed at five time points: 48, 96, 144, 192, and 240 h after the harvest.

STATISTICAL A^NALYSIS

With reference to the calibration procedure, multivariate linear regression was performed for each analyzed parameter on the data from the 1800 fruits that were collected as described above.

To study the processing plant and validate the use of the vis/NIR instrument, our factorial experiment evaluated the changes in three variables (firmness, pH, and SSC of mandarins) for different levels of two factors (test and time). A balanced factorial experiment was chosen in which five observations (fruit samples) were randomly taken for each combination of the levels of the two factors. The test factor consisted of the four sampling points (A, B, C, and D, as shown in table 1), and the time factor was the five time points (48, 96, 144, 196, and 240 h after harvest, as described above). Thus, a 4 × 5 × 5 factorial experiment was considered.

Our interest was to verify the existence of significant effects of the considered factors on the parameters of interest.

In other words, we were interested in testing whether different levels of factors have equal means. The significance of treatment effects was evaluated with analysis of variance (ANOVA; Agresti and Finlay, 2012). The F-test was considered to evaluate the significance of treatment (F = MSTR/MSE, where MSTR is the sum of squares of the treatments, and MSE is the sum of squares of the error).

When a treatment is significant, the means significantly differ. We tested:

4 1

0:

H α ==α

5 1

0:

H γ ==γ

( ) ( )

11 12

( )

45 0:

H αγ = αγ == αγ

Figure 2. The vis/NIR spectrophotometer (NCS001, Sacmi, Imola, Italy) that was used in the tests.

(4)

where α_i, γ_j, and

( )

αγij denote the means of different levels of the factors test, time, and test × time, respectively.

Linear contrasts were calculated to determine which pairs of means significantly differed within the levels of each factor. To adjust the p-values for multiple testing, the Bonferroni method was considered. Tests of interactions indicate whether the cell means for the interaction are equal. Tests of main effects indicate whether the marginal cell means for the factor are equal. The contrasts were calculated by comparing each conditional mean with both the mean of a reference category and the adjacent mean (each mean level with the subsequent level).

In the second step, the accuracy of the vis/NIR device was evaluated. Based on the factorial design and ANOVA model results, the significance of the linear correlation coefficients for each parameter of interest was evaluated by estimating a regression model and checking the significance of the parameters with t-tests. Our estimates are model-based:

For firmness, the estimated model is:

ijk j i ij ijk

yijk =α+β₁NIR +β Test ⋅Time +ε (1) For pH, the estimated model is:

ijk ijk

yijk =α+β₁NIR +ε (2)

For SSC, the estimated model is:

ijk j j i i ijk

yijk =α+β₁NIR +βTest +β Time +ε (3) where

yijk = kth observation of the parameter of interest, which corresponds to test level i, time level j, and k = 1 to 5 NIRijk = data predicted using the vis/NIR device Testi = four levels of the test factor (i = 1 to 4) Timej = five levels of the time factor (j = 1 to 5) Testi⋅ Timej = i × j interaction factors

εijk = usual random error component with μ = 0 and variance σ².

The significance of the partial correlation between the observed and predicted values for different levels of the time and test factors was evaluated with t-tests.

R

ESULTS AND

D

ISCUSSION VIS/NIRDEVICE CALIBRATION

Table 3 shows the statistical evaluation results for the validation of the portable vis/NIR device calibration models. It reveals the prediction values of firmness, pH, and SSC by vis/NIR spectroscopy, which are similar to those obtained in the laboratory by direct measurement of these parameters; thus, the application of these models is feasi-

ble. In particular, the bias prediction values are lower than the bias calibration values for the three quality attributes.

This reveals the good ability of the instrument in predicting these attributes.

D^ATAS^{ET FROM}DESTRUCTIVE ANALYSES OF F^RUITS The descriptive statistics for fruit firmness, pH, and SSC are reported in table 4. These parameters were derived from the destructive tests on 100 fruits at different points in the processing plant (test factor) and at different time points after harvest (time factor). Firmness shows the highest variability (CV = 0.223), followed by SSC (CV = 0.11).

FIRMNESS

The compression tests gave interesting information about fruit firmness at different sampling points in the processing plant during the study period of 10 days after harvest (fig. 3). Gomez et al. (2006) obtained 55.2 N as the maximum compression force on Satsuma mandarins, but they associated that value with a 3% reduction in the height diameter, whereas we considered the peak force or maximum force that was recorded during penetration, as Beghi et al. (2014) did with apples.

Table 5 shows the ANOVA results for firmness, which was evaluated by destructive tests. The interaction effects and main effects of test (A, B, C, and D) and time (48, 96,

Table 3. Statistics of calibration and prediction.^[a]

Variable

Calibration Prediction Obs. SEC Bias Obs. SEP Bias Firmness (N) 1800 3.25 -4.10 100 2.840 -3.11

pH 1800 0.106 -0.20 100 0.016 5 e^-7 SSC (°Brix) 1800 0.160 -0.10 100 0.126 4 e^-8

[a] Obs. = number of samples, SEC = standard error of calibration, and SEP = standard error of prediction.

Table 4. Descriptive statistics of fruit firmness, pH, and SSC from destructive analyses.

Variable Obs. Mean SD Min. Max. CV Firmness (N) 100 127.113 28.286 69 177 0.223

pH 100 4.000 0.156 3.4 4.32 0.039 SSC (°Brix) 100 11.612 1.278 8.95 15.8 0.110

Figure 3. Fruit firmness by destructive tests. Fruits were taken at different sampling points of the processing plant (A, B, C, and D) at different time points after harvest (48, 96, 144, 196, and 240 h). The data are reported as means of five replicates ±SD.

Table 5. ANOVA results for fruit firmness.^[a]

Source Partial SS df MS F Prob>F Model 76264.846 19 4013.939 109.08 0.000

Time 60347.634 4 15086.909 410.01 0.000 Test 14189.377 3 4729.792 128.54 0.000 Time _× Test 1727.834 12 143.986 3.91 0.000

Error 2943.723 80 36.797 Total 79208.568 99 800.086

[a] R² = 0.9628; adjusted R² = 0.9540.

0 20 40 60 80 100 120 140 160 180 200

A B C D

Firmness [N]

Test

48 h 96 h 144 h 192 h 240 h

(5)

144, 196, and 240 h after the harvest) are highly significant.

The contrasts of the adjusted prediction of the test and time cross effect are shown in figure 4. These contrasts are the joint tests of interactions of time and test within each contrast, which is defined by the adjusted mean of the reference category (A). The first point on the x-axis is the comparison between the means of test B with respect to A (B vs. A) for different levels of time (48, 96, 144, 192, and 240 h); there is no significant difference between the means in different values of time at this point. It is tempting to conclude from this overlap that the differences are not statistically significant. However, there are statistically significant differences between test levels C and D compared to A at higher time levels (144, 192, and 240 h) with respect to 48 and 96 h.

Figure 5 shows the average profile of the linear prediction of the test levels for different time levels. There is no significant difference between the means of test at the first two levels of time; conversely, the means of tests A and B differ from those of tests C and D. Note that the difference between A and B increases with the time level, which indi- cates statistically significant differences in fruit firmness at points A and B of the processing plant far from harvest. In other words, as fruit deterioration increases, the firmness has a higher average effect in absolute terms, that is, the index of the damage caused by processing on fruits.

A regression model that considers the factorial design and the results of ANOVA (model 1) were estimated to evaluate the performance of the vis/NIR device. The significance of the parameters and the vis/NIR device accuracy were evaluated from the coefficients of the estimated model shown in equation 1. The results (table 6) show that the vis/NIR device can predict 96% of the total variability of the observed values. In addition, the insignificance of the coefficients that correspond to different test levels (at different sampling points of the processing plant) in the early stages of time denotes the poor ability of the instrument to properly detect the firmness values in the time points closer to the fruit processing. This result can be considered a limit of the vis/NIR device and could be explained by the fact that mandarins are intact or little damaged at the beginning of fruit processing (tests A and B), and the instrument is not able to identify possible damage at a low level in terms of fruit firmness, and the same with reference to the variable time. Conversely, the effect of the parameters increases with increasing levels of test and time. The random distribution of residuals in model 1 for fruit firmness (fig. 6) is

Figure 4. Contrasts of adjusted predictions of cross test-time levels.

Figure 5. Profile plot of adjusted predictions of time × test.

Table 6. Parameter estimates of model 1 for fruit firmness to test the prediction performance of the vis/NIR device.^[a]

Coefficient SE t P>t α 109.13 9.04 12.07 0 NIR 0.36 0.06 6.25 0 48h#B -0.02 0.02 -0.65 0.52 48h#C -0.03 0.03 -0.95 0.34 48h#D -0.05 0.03 -1.6 0.11 96h#A -0.05 0.02 -2.13 0.04 96h#B -0.09 0.02 -3.83 0 96h#C -0.12 0.02 -5.13 0 96h#D -0.17 0.03 -5.97 0 144h#A -0.08 0.03 -3.02 0 144h#B -0.13 0.03 -4.48 0 144h#C -0.26 0.04 -7.25 0 144h#D -0.3 0.04 -7.04 0 192h#A -0.2 0.03 -6.49 0 192h#B -0.3 0.03 -8.8 0 192h#C -0.47 0.05 -10.05 0 192h#D -0.51 0.05 -10.03 0 240h#A -0.3 0.04 -6.94 0 240h#B -0.48 0.05 -10.02 0 240h#C -0.71 0.07 -10.34 0 240h#D -0.88 0.08 -11.23 0

[a] R² = 0.9649.

Figure 6. Residual distribution in model 1 for fruit firmness.

(6)

consistent with the assumptions of normality.

Bureau et al. (2009) determined that a vis/NIR prediction model for apricot firmness was not satisfactorily accurate because of the high calibration and prediction errors.

However, Gomez et al. (2006) found an acceptable correlation (r = 0.83) between NIR measurements and destructive test results for mandarin firmness with high standard errors of calibration and prediction (8.18 and 8.62 N, respectively) due to the large standard deviation of the data sets. In Cayuela and Weiland (2010), the ability of NIR to estimate orange flesh firmness reached accuracy levels of 83.9% and 79.0%, respectively, with the two portable NIR spectrometers (Labspec and Luminar 5030). The difficulty in corre- lating destructive and nondestructive measurements of intact orange firmness due to the thick peel of the fruits was pointed out by Sanchez et al. (2013); they obtained an un- acceptable prediction performance (maximum r² = 0.33).

An acceptable prediction model for apple firmness during storage was obtained by Beghi et al. (2014) (r² = 0.83 for Golden Delicious variety).

PHVALUES

The pH values of the fruits at the different sampling points of the processing plant during the entire study period (fig. 7) are comparable with those obtained by Gomez et al.

(2006). The ANOVA results for pH (table 7) show that both the effect of interaction and the main effects are not significant; thus, no averages significantly differ. The ANOVA model leads us to state that there is no significant difference among the group means. From these considera- tions, we decided to use model 2 to evaluate the correlation between measured and predicted pH values.

The results (table 8) show the excellent ability of the instrument in predicting pH values (r² = 0.93), which con- firms the results of Gomez et al. (2006) for Satsuma mandarin (r = 0.87). Each pH unit is measured by the instrument for 0.95. Overall, the vis/NIR device explains 93% of the variability of the observed pH. This result can be linked to the low variability of the pH parameter. The residuals show a random distribution of errors (fig. 8), and even some underestimated pH values can be noted. Nonetheless, vis/NIR application on intact oranges for pH and acidity estimation gave limited results (r² = 0.45) due to the rela- tively low levels of organic acids in oranges with respect to soluble solids concentrations (Cayuela, 2008). A better performance was obtained for pH evaluation of intact oranges by Cayuela and Weiland (2010) (r = 0.90). The pH prediction model reported by Sanchez et al. (2013) for intact oranges was not suitable for routine applications (r² = 0.10 to 0.32) because of the considerable uniformity of the data set used; in this case, the fruits were harvested at commercial maturity.

SSCVALUES

The obtained SSC values (fig. 9) are within the range that other authors found for mandarin (Gomez et al., 2006;

Sun et al., 2009; Liu et al., 2010b; Antonucci et al., 2011) and orange (Cayuela, 2008; Cayuela and Weiland, 2010;

Liu et al., 2010a; Sanchez et al., 2013).

The complete ANOVA model showed that the interactions were not significant; therefore, the model without interactions was estimated (table 9). Although the main effects are significant, they only explain 54% of the total variability. In addition, the existence of significant differences for the means of the main effects test (table 10) and time (table 11) was evaluated.

The means of the levels of each factor with respect to the subsequent levels are reported in tables 10 and 11. Note that test A significantly differs from the subsequent tests,

Figure 7. pH values by destructive tests. The fruits were obtained at different sampling points of the processing plant (A, B, C, and D) at different time points after harvest (48, 96, 144, 196, and 240 h). The data are reported as means of five replicates ±SD.

Table 7. ANOVA results for fruit pH.^[a]

Time 0.058 4 0.015 0.620 0.652 Test 0.063 3 0.021 0.448 0.890 Test _× Time 0.409 12 0.034 1.440 0.164

Error 1.887 80 0.024 Total 2.418 99 0.024

[a] R² = 0.219; adjusted R² = 0.034.

Table 8. Parameter estimates of model 2 for fruit pH to test the performance of the vis/NIR device.^[a]

Coefficient SE t P>t NIR 0.95 0.03 36.09 0.00

α 0.21 0.11 2.02 0.05

[a] R² = 0.93; adjusted R² = 0.93.

Figure 8. Residual distribution of model 2 for pH.

3.0 3.5 4.0 4.5 5.0

A B C D

pH

Test

48 h 96 h 144 h 192 h 240 h

(7)

i.e., B, C, and D. With reference to time, the mean SSC values for the first three time points (48, 96, and 144 h) significantly differ from those of the subsequent time points. No difference was found between the SSC values at 192 and 240 h.

Finally, the linear model that correlates the observed and predicted values considering the main time and test effects based on the ANOVA results was considered to evaluate the correlation between the SSC measurements from destructive tests with those predicted by the vis/NIR device, i.e., the prediction ability of the instrument. The vis/NIR device can explain 93% of the total SSC variability (table 12). The mean SSC values of test levels B and C are significantly different; for the time factor, significant differences were obtained between the highest levels (192 and 240 h). The low values of the parameters for different lev-

els of the factors encourage us to verify the significance of the reduced model (F(8;91)≈ 2.02; p = 0.01), which contin- ues to explain a variability of 0.93 and shows an excellent ability of the vis/NIR device in predicting SSC values. Fur- thermore, the notably low standard error emphasizes the efficiency of the estimator. Finally, the residual plot (fig. 10) shows the relationships of the assumed random errors and normality.

Gomez et al. (2006) also obtained significantly accurate SSC evaluation using vis/NIR spectroscopy on Satsuma mandarin (r = 0.94; SEP = 0.32). SSC prediction was also good in Sun et al. (2009) and Liu et al. (2010b), with r = 0.92 and SEP 0.65, demonstrating that vis/NIR is a feasible method to nondestructively measure SSC in Nanfeng mandarin. Similar regression values were obtained for the SSC evaluation of other fruits (Beghi et al., 2014; Bureau et al., 2009; Liu and Ying, 2005), particularly oranges. Different models constructed with several spectrometers by different authors were generally able to predict SSC in oranges with an accuracy higher than 90% (Liu et al., 2010a; Cayuela and Weiland, 2010).

C

ONCLUSIONS

As stated by many studies, portable vis/NIR instruments are useful to rapidly and non-destructively evaluate the quality of fruits and vegetables. This study represents a new contribution concerning the use of a portable vis/NIR system on mandarins that were processed on an industrial scale to identify possible damage to fruits during processing before packaging. The non-destructive evaluation of some qualitative parameters of “Tardivo di Ciaculli”

mandarin using the vis/NIR technology was extended to a period of 10 days after harvest.

Using the vis/NIR device to predict mandarin firmness,

Figure 9. SSC values by destructive tests. The fruits were obtained at different sampling points of the processing plant (A, B, C, and D) at different time points after harvest (48, 96, 144, 196, and 240 h). The data are reported as means of five replicates ±SD.

Table 9. ANOVA results for SSC (number of observations = 100).^[a]

Time 63.496 4 15.874 19.87 0.000 Test 24.498 3 8.166 10.22 0.000 Error 73.516 92 0.799

Total 161.511 99 1.631

[a] R² = 0.545; adjusted R² = 0.510.

Table 10. Significance of linear contrasts of test means (p-value adjusted by Bonferroni method). Each mean was compared with the subsequent means.

Test df F P>F

A vs. >A 1 27.48 0 B vs. >B 1 2.21 0.1407

C vs. D 1 0.97 0.3269

Joint 3 10.22 0

Denominator 92

Table 11. Significance of linear contrasts of time means (p-value was adjusted using Bonferroni method). Each mean was compared with the subsequent means.

Time df F P>F

48 h vs. >48 h 1 21.64 0 (96 h vs. >96 h 1 42.73 0 144 h vs. >144 h 1 13.89 0.000

192 h vs. 240 h 1 1.21 0.273

Joint 4 19.87 0

Figure 10. Residual distribution of model 3 for SSC.

Table 12. Parameter estimates of model 3 for SSC to test the prediction performance of the vis/NIR device (number of observations

= 100).^[a]

Coefficient SE t P>t 95% CI NIR 0.97 0.03 35.83 0.00 0.92-1.03

α 0.28 0.32 0.87 0.39 -0.35-0.91

[a] R² = 0.93; adjusted R² = 0.93.

6 8 10 12 14 16

A B C D

SSC [°Brix]

Test

48 h 96 h 144 h 192 h 240 h

(8)

the results showed a poor prediction ability at time points near harvest and insignificant test levels (different sampling points in the processing plant) at the first time levels, that is, closer to harvest. The pH estimation of “Tardivo di Ciaculli” mandarin was excellent at different sampling points and in time, probably because of the small variability in pH. The vis/NIR device also estimated SSC excellently, which explains 93% of the total variability of the parameter.

For the SSC values, the results indicate statistically significant differences between harvest and all subsequent sampling points in the processing plant. For the effect of time on SSC, statistically significant differences were obtained only for the first three time levels (48, 96, and 144 h after harvest) compared with the subsequent means. The mean SSC values at sampling points B and C (i.e., fruit unloading and manual selection) and for the highest time levels (192 and 240 h) were significantly different.

Therefore, it can be assumed that vis/NIR technology (600-1000 nm) can be applied to “Tardivo di Ciaculli”

mandarin to precisely estimate the pH and SSC values. For fruit firmness, the estimation appears to be insufficiently accurate, particularly at time steps closer to fruit harvest and in the evaluation of possible mechanical damage during the processing stages on an industrial scale.

Recalibration may be required in further research, with particular reference to fruit firmness, to implement the use of NIR spectroscopy directly in the processing plant. Fur- ther research would probably need to consider other indica- tors of mandarin texture, rather than firmness, in evaluating the potential damage to fruit.

ACKNOWLEDGEMENTS

The authors are grateful to Dr. Giovanni D’Agati, presi- dent of the “Consorzio del Mandarino Tardivo di Ciaculli”

in Palermo, Italy, for providing the fruits and access to the processing plant, and to Mr. Salvatore Amoroso for sup- porting the test execution. The Vis/NIR system (NCS001 by Sacmi, Imola, Italy) was purchased within the project

“PLASS” PONa3_00053.

R

EFERENCES

Agresti, A., & Finlay, B. (2012). Metodi statistici di base e avanzati per le scienze sociali. Milan, Italy: Pearson Italia.

Antonucci, F., Pallottino, F., Paglia, G., Palma, A., D’Aquino, S., &

Menesatti, P. (2011). Non-destructive estimation of mandarin maturity status through portable VIS-NIR spectrophotometer.

Food Bioproc. Tech., 4(5), 809-813.

http://dx.doi.org/10.1007/s11947-010-0414-5

Beghi, R., Giovanelli, G., Malegori, C., Giovenzana, V., & Guidetti, R. (2014). Testing of a VIS-NIR system for the monitoring of long-term apple storage. Food Bioproc. Tech., 7(7), 2134-2143.

http://dx.doi.org/10.1007/s11947-014-1294-x

Beghi, R., Spinardi, A., Bodria, L., Mignani, I., & Guidetti, R.

(2012). Apples nutraceutical properties evaluation through a visible and near-infrared portable system. Food Bioproc. Tech., 6(9), 2547-2554. http://dx.doi.org/10.1007/s11947-012-0824-7 Bureau, S., Ruiz, D., Reich, M., Gouble, B., Bertrand, D.,

Audergon, J. M., & Renard, C. M. G. C. (2009). Rapid and nondestructive analysis of apricot fruit quality using FT-near-

infrared spectroscopy. Food Chem., 113(4), 1323-1328.

http://dx.doi.org/10.1016/j.foodchem.2008.08.066

Catania, P., Alleri, M., Martorana, A., Settanni, L., Moschetti, G., &

Vallone, M. (2014). Investigation on a tunnel pasteurizer for Nocellara del Belice table olives according to the Castelvetrano method. Grasas Aceites, 65(4).

http://dx.doi.org/10.3989/gya.0578141

Catania, P., Vallone, M., Planeta, D., & Febo, P. (2015).

Instrumental evaluation of the texture of cv. Nocellara del Belice table olives. Rivista Italiana Sostanze Grasse, 92(1), 43-51.

Cayuela, J. A. (2008). Vis/NIR soluble solids prediction in intact oranges (Citrus sinensis L.) cv. Valencia Late by reflectance.

Postharvest Biol. Tech., 47(1), 75-80.

http://dx.doi.org/10.1016/j.postharvbio.2007.06.005 Cayuela, J. A., & Weiland, C. (2010). Intact orange quality

prediction with two portable NIR spectrometers. Postharvest Biol. Tech., 58(2), 113-120.

http://dx.doi.org/10.1016/j.postharvbio.2010.06.001 Fraser, D. G., Jordan, R. B., Kunnemeyer, R., & McGlone, V. A.

(2003). Light distribution inside mandarin fruit during internal quality assessment by NIR spectroscopy. Postharvest Biol.

Tech., 27(2), 185-196. http://dx.doi.org/10.1016/S0925- 5214(02)00058-3

Giovanelli, G., Sinelli, N., Beghi, R., Guidetti, R., & Casiraghi, E.

(2014). NIR spectroscopy for the optimization of postharvest apple management. Postharvest Biol. Tech., 87, 13-20.

http://dx.doi.org/10.1016/j.postharvbio.2013.07.041 Gomez, A. H., He, Y., & Pereira, A. G. (2006). Non-destructive

measurement of acidity, soluble solids, and firmness of Satsuma mandarin using Vis/NIR-spectroscopy techniques. J. Food Eng., 77(2), 313-319.

http://dx.doi.org/10.1016/j.jfoodeng.2005.06.036 Guthrie, J. A., Walsh, K. B., Reid, D. J., & Liebenberg, C. J.

(2005a). Assessment of internal quality attributes of mandarin fruit: 1. NIR calibration model development. Australian J. Agric.

Res., 56(4), 405-416. http://dx.doi.org/10.1071/AR04257 Guthrie, J. A., Reid, D. J., & Walsh, K. (2005b). Assessment of

internal quality attributes of mandarin fruit: 2. NIR calibration model robustness. Australian J. Agric. Res., 56(4), 417-426.

http://dx.doi.org/10.1071/AR04299

Kawano, S., Fujiwara, K., & Iwamoto, M. (1993). Nondestructive determination of sugar content in Satsuma mandarin using near- infrared transmittance. J. Japanese Soc. Hort. Sci., 62, 465-470.

http://dx.doi.org/10.2503/jjshs.62.465

Lammertyn, J., Nicolai, B., Ooms, K., De Smedt, V., & De Baerdemaeker, J. (1998). Non-destructive measurement of acidity, soluble solids, and firmness of Jonagold apples using NIR spectroscopy. Trans. ASAE, 41(4), 1086-1094.

http://dx.doi.org/10.13031/2013.17238

Liu, Y., & Ying, Y. (2005). Use of FT-NIR spectrometry in non- invasive measurements of internal quality of Fuji apples.

Postharvest Biol. Tech., 37(1), 65-71.

http://dx.doi.org/10.1016/j.postharvbio.2005.02.013 Liu, Y., Sun, X., & Aiguo, O. (2010a). Nondestructive

measurement of soluble solids content of navel orange fruit by visible-NIR spectrometric technique with PLSR and PCA- BPNN. Food Sci. Tech., 43(4), 602-607.

http://dx.doi.org/10.1016/j.lwt.2009.10.008

Liu, Y., Sun, X., Zhang, H., & Aiguo, O. (2010b). Nondestructive measurement of internal quality of Nanfeng mandarin fruit by charge-coupled device near-infrared spectroscopy. Comput.

Electron. Agric., 71(supp. 1), 510-514.

http://dx.doi.org/10.1016/j.compag.2009.09.005

Lu, H., Jiang, H., Fu, X., Yu, H., Xu, H., & Ying, Y. (2008). Non- invasive measurements of the internal quality of intact Gannan navel orange by VIS/NIR spectroscopy. Trans. ASABE, 51(3),

(9)

1009-1014. http://dx.doi.org/10.13031/2013.24505 Magwaza, L. S., Opara, U. L., Cronje, P. J. R., Landahl, S.,

Nieuwoudt, H. H., Mouazen, A. M., Nicolai, B. M., & Terry, L.

A. (2014). Assessment of rind quality of Nules Clementine mandarin during postharvest storage: 1. Vis/NIRS PCA models and relationship with canopy position. Sci. Hort., 165, 410-420.

http://dx.doi.org/10.1016/j.scienta.2013.09.035

Magwaza, L. S., Opara, U., Nieuwoudt, E., Cronje, P. J. R., Saeys, W., & Nicolaï, B. (2012). NIR spectroscopy applications for internal and external quality analysis of citrus fruit: A review.

http://dx.doi.org/10.1007/s11947-011-0697-1

McGlone, V. A., Fraser, D. G., Jordan, R. B., & Künnemeyer, R.

(2003). Internal quality assessment of mandarin fruit by VIS/NIR spectroscopy. J. Near Infrared Spectroscopy, 11(5), 323-332. http://dx.doi.org/10.1255/jnirs.383

Menesatti, P., Zanella, A., D’Andrea, S., Costa, C., Paglia, G., &

Pallottino, F. (2009). Supervised multivariate analysis of hyperspectral NIR images to evaluate the starch index of apples.

http://dx.doi.org/10.1007/s11947-008-0120-8

Miller, W. M., & Zude-Sasse, M. (2004). NIR-based sensing to measure soluble solids content of Florida citrus. Appl. Eng.

Agric., 20(3), 321-327. http://dx.doi.org/10.13031/2013.16052 Nicolai, B. M., Beullens, K., Bobelyn, E., Ann Peirs, A., Saeys, W.,

Theron, K. I., & Lammertyn, J. (2007). Nondestructive measurement of fruit and vegetable quality by means of NIR spectroscopy: A review. Postharvest Biol. Tech., 46(2), 99-118.

http://dx.doi.org/10.1016/j.postharvbio.2007.06.024

Ortiz, C., & Torregrosa, A. (2014). Mechanical properties of citrus and impact damage under different storage conditions. Trans.

ASABE, 57(2), 593-598.

http://dx.doi.org/10.13031/trans.57.10479

Osborne, S. D., Kunnemeyer, R., & Jordan, R. B. (1999). A low- cost system for the grading of kiwifruit. J. Near Infrared

Spectros., 7(1), 9-15. http://dx.doi.org/10.1255/jnirs.229 Park, B., Abbott, J. A., Lee, K. J., Choi, C. H., & Choi, K. H.

(2003). Near-infrared diffuse reflectance for quantitative and qualitative measurement of soluble solids and firmness of delicious and gala apples. Trans. ASAE, 46(6), 1721-1731.

http://dx.doi.org/10.13031/2013.15628

Sánchez, M. T., De la Haba, M. J., Serrano, I., & Pérez-Marín, D.

(2013). Application of NIRS for nondestructive measurement of quality parameters in intact oranges during on-tree ripening and at harvest. Food Anal. Methods, 6(3), 826-837.

http://dx.doi.org/10.1007/s12161-012-9490-7

Saranwong, S., Sornsrivichai, J., & Kawano, S. (2004). Prediction of ripe-stage eating quality of mango fruit from its harvest quality measured nondestructively by near-infrared spectroscopy. Postharvest Biol. Tech., 31(2), 137-145.

http://dx.doi.org/10.1016/j.postharvbio.2003.08.007

Slaughter, D. C., Barrett, D., & Boersig, M. (1996). Nondestructive determination of soluble solids in tomatoes using near-infrared spectroscopy. J. Food Sci., 61(4), 695-697.

http://dx.doi.org/10.1111/j.1365-2621.1996.tb12183.x Steuer, B., Schulz, H., & Làger, E. (2001). Classification and

analysis of citrus oils by NIR spectroscopy. Food Chem., 72(1), 113-117. http://dx.doi.org/10.1016/S0308-8146(00)00209-0 Sun, X., Zhang, H., & Liu, Y. (2009). Nondestructive assessment of

quality of Nanfeng mandarin fruit by portable near-infrared spectroscopy. Intl. J. Agric. Biol. Eng., 2(1), 65-71.

Zheng, Y., He, S., Yi, S., Zhou, Z., Mao, S., Zhao, X., & Deng, L.

(2010). Predicting oleocellosis sensitivity in citrus using VNIR reflectance spectroscopy. Sci. Hort., 125(3), 401-405.

http://dx.doi.org/10.1016/j.scienta.2010.04.036

Zude, M., Pflanz, M., Kaprielian, C., & Aivazian, B. L. (2008).

NIRS as a tool for precision horticulture in the citrus industry.

Biosyst. Eng., 99(3), 455-459.

http://dx.doi.org/10.1016/j.biosystemseng.2007.10.016