Microsoft Word - 54catania.docx
CHEMICAL ENGINEERING TRANSACTIONS
VOL. 44, 2015
A publication of
The Italian Association
of Chemical Engineering
Online at www.aidic.it/cet
Guest Editors:Riccardo Guidetti,Luigi Bodria, Stanley Best
Copyright © 2015, AIDIC ServiziS.r.l.,
ISBN 978-88-95608-35-8;ISSN 2283-9216
Time-resolved Reflectance Spectroscopy as a Management
Tool for Late-maturing Nectarine Supply Chain
Anna Rizzolo*a, Maristella Vanolia,b, Maurizio Grassia, Lorenzo Spinellib,
Alessandro Torricellic
aConsiglio per la ricerca in agricoltura e l’analisi dell’economia agraria (CRA) – Unità di ricerca per i processi dell’industria
agroalimentare (CRA-IAA), via G. Venezian 26, Milano (Italy)]
bIstituto di Fotonica e Nanotecnologie – CNR, Piazza Leonardo da Vinci 32, Milano (Italy)
cPolitecnico di Milano, Dipartimento di Fisica, Piazza Leonardo da Vinci 32, Milano (Italy)
anna.rizzolo@entecra.it
The absorption coefficient of the fruit flesh at 670 nm (μa), measured at harvest by time-resolved reflectance
spectroscopy (TRS) is a good maturity index for early nectarine cultivars. A kinetic model has been developed
linking the μa, expressed as the biological shift factor to softening during ripening. This allows shelf life
prediction for individual fruit from the value of μa at harvest and the fruit categorization into predicted softening
and usability classes. In this work, the predictive capacity of a kinetic model developed using μa data at
harvest and firmness data within 1-2 d after harvest for a late maturing nectarine cultivar (‘Morsiani 90’) was
tested for prediction and classification ability. Compared to early maturing cultivars, μa at harvest had low
values and low variability, indicating advanced maturity, whereas firmness was similar. Hence, fruit were
categorized into six usability classes (from ‘transportable-hard’ to ‘ready-to-eat-very soft’) basing on μa limits
established analyzing firmness data in shelf life after harvest. The model was tested by comparing the
predicted firmness and class of usability to the actual ones measured during ripening and its performance
compared to that of models based on data during the whole shelf life at 20 °C after harvest and after storage
at 0 °C and 4 °C. The model showed a classification ability very close to that of models based on data of the
whole shelf life, and was able to correctly segregate the ‘ready-to-eat-transportable’, ‘transportable’ and
‘transportable-hard’ classes for ripening at harvest and after storage at 0 °C, and the ‘ready-to-eat-very soft’
and ‘ready-to-eat-soft’ classes for ripening after storage at 4 °C, with lower performance of models for fruit
after storage at 4 °C respect to those of the other two ripening.
1. Introduction
It is wellknown that peach and nectarine fruit quality is strictly dependent on its maturity and that there is a
large variation in maturity, even within the same harvest date, which could have an impact during subsequent
marketing and consumption. In fact, if harvested too early, they lack flavor, and sometimes, ripening capacity
whereas when harvested ripe, they have excellent eating quality but may be subjected to mechanical injury
and decay during handling (Crisosto and Valero, 2008). Time-resolved reflectance spectroscopy (TRS) is a
non-destructive technique based on the injection of a short pulse of monochromatic light in the fruit flesh down
to a 1–2 cm in depth from fruit surface and on the analysis of time distribution of re-emitted photons, allowing
the differentiation between the absorption coefficient (µa), related to chemical composition, and the reduced
scattering coefficient (µ’s), related to physical structure (Cubeddu et al., 2001).
Previous research has shown that maturity of fruit at harvest can be assessed non-destructively by using TRS
to measure the absorption coefficient of the fruit flesh at 670 nm (μa), near the chlorophyll peak. In nectarines,
as fruit maturation and ripening proceed, the μa value decreases following a logistic curve (Tijskens et al.,
2006) and is synchronized with softening (Tijskens et al., 2007). Hence a kinetic model has been developed
linking the μa, expressed as the biological shift factor (BSF), to softening during ripening, so including the
variations in maturity at harvest in the firmness decay model. In this way from the value of μa at harvest, the
DOI: 10.3303/CET1544002
Please cite this article as: Rizzolo A., Vanoli M., Grassi M., Spinelli L., Torricelli A., 2015, Time-resolved reflectance spectroscopy as a
management tool in the fruit supply chain, Chemical Engineering Transactions, 44, 7-12 DOI: 10.3303/CET1544002
7
shelf life for individual fruit can be predicted. In order to validate this methodology and to evaluate the
predictive capacity of the kinetic model for an early maturing nectarine cultivar, Rizzolo et al. (2009)
segregated ‘Spring Bright’ fruit according to their softening capacity (‘will never soften’, ‘dangerously hard’,
‘transportable’, ‘ready to eat-firm’, ‘ready to eat-ripe’ and ‘overripe’) on the basis of the value of μa at harvest.
With an export trial from Italy to The Netherlands, simulating on a small scale (1000 fruit) the fruit supply chain
from the packing-house to the consumer, Eccher Zerbini et al. (2009) showed that ripening classes had been
correctly predicted. Applying this methodology at the time of harvest, Rizzolo et al. (2009) found for the early
maturing ‘Spring Bright’ nectarines that measuring μa on all fruit and firmness on two samples of about 30 fruit,
representative of all the μa range, the first as soon as possible after harvest and the second after 24 h at
20 °C, it was possible to estimate the parameters of the firmness decay model for the season and cultivar, and
hence to compute the time required to reach the midpoint of the firmness decay curve of the μa values in each
softening class. Then, these time values were used to select fruit with different stages of maturity for different
marketing segments, such as distant or close-by markets.
In this work, the predictive capacity of a kinetic model developed for a late maturing nectarine cultivar
(‘Morsiani 90’) by Eccher Zerbini et al. (2011) by using μa data at harvest and firmness data within 1-2 d after
harvest was tested for prediction and classification ability compared to that of models based on data during
the whole shelf-life at 20 °C after harvest and after storage at 0 °C and 4 °C.
2. Materials and Methods
2.1 Fruit and Experimental plan
In season 2009, ‘Morsiani 90’ nectarines were picked in Faenza (Italy) at the commercial harvest. The details
of the experimental plan have been described by Eccher Zerbini et al. (2011) and Lurie at al. (2011). In this
work, fruit assigned for shelf life at 20 °C at harvest as well as those assigned for shelf life at 20 °C after 4
weeks storage at 0 °C and 4 °C were considered. The day after harvest nectarines with defects and bruises
were removed, and the resulting fruit were individually measured by TRS at 670 nm using a prototype built at
Politecnico di Milano (Torricelli et al., 2008) and then ranked by decreasing μa value. The ranked fruit were
grouped by 16, with a total of 30 groups, corresponding to 30 levels of μa. Each fruit from each group was
randomly assigned to a different sample. In this way, 16 samples were obtained each one containing 30 fruit
from the whole range of μa and dedicated to one time of analysis according to Table 1.
Table 1: Samples and times (h) of shelf life for firmness measurements
Shelf life Samples Shelf life times (h)
harvest 0-5 29, 55, 74, 101, 175, 198
after storage at 0 °C 6-10 37, 60, 80, 108, 131
after storage at 4 °C 11-15 37, 60, 80, 108, 131
Firmness (F) was measured by a penetrometer (Texture Analyzer TA.XtPlus, Stable Micro Systems, England,
8 mm diameter plunger, crosshead speed 3.33 mm s−1) on opposite sides of each fruit after skin removal
approximately on the same spot where also μa had been measured.
2.2 Data processing
The μa values of individual fruit were converted into the BSF (∆t*µa) according to the equation developed by
Tijskens et al. (2006), and the BSF relative to firmness curve (∆t*F) was computed according to Tijskens et al.
(2005). In nectarines, the BSF for μa and the BSF for firmness are linearly related (Tijskens et al., 2007)
according to Eq(1), where α and β are parameters to be estimated.
∆t*F = α(∆t*µa + β) (1)
The firmness decay model has been described by Tijskens et al. (2007) and is reported in Eq(2):
(2)
where Fmax is the maximum firmness at minus infinite time, Fmin is the minimum firmness achieved at infinite
time, kf is the softening rate constant at 20 °C, t is time, ∆t*F is the BFS for firmness.
The parameters of firmness decay model for ripening at 20 °C at harvest, and after storage at 0 °C and 4 °C
have been presented and discussed by Eccher Zerbini et al. (2011) and Table 2 summarizes the parameters
of the models tested in this work for prediction and classification ability.
= − 1 + ( − ) +∆ ∗ +
8
Table 2: Parameters of the non-linear regression model for firmness decay estimated in ‘Morsiani 90’
nectarines kept at 20 °C after harvest or after storage at 0 °C and 4 °C (Eccher Zerbini et al., 2011)
Model Code Samples Fmax Fmin α β kf20°C
At harvest + 24 h H+24h 0-1 77 3.5 1.61 −2.44 0.00031
Ripening after harvest T harvest 0-5 85 4.7 1.40 −2.27 0.000234
After storage at 0 °C T 0C 6-10 85 4.7 1.03 −2.72 0.000227
After storage at 4 °C T 4C 11-15 85 4.7 0.70 −2.31 0.000340
2.3 Prediction ability of firmness decay models
The predicted firmness (Fpred) of every fruit during shelf-life, computed from Eqs (1) and (2), was compared to
the measured firmness (Fmeas) by using linear regression analysis. The mean absolute error (MAE), the
average deviation (AD), the mean square error (MSE), the root mean standard error of deviation (RMSED)
and the ratio of the standard deviation of Fmeas to RMSED (s/RMSED) were chosen to measure the
performance of models and were computed according to the equations reported by Rizzolo et al. (2009).
2.4 Misclassification of models
Fruit were categorized into six µa classes of predicted firmness potential for handling and eating (Mi) according
to the µa limits described and discussed in Results and Discussion, corresponding to different uses and/or
softening potentials. Hence, every model was tested for misclassification according to the criteria reported by
Rizzolo et al. (2009): in each class Mi based on the µa value at harvest, the Fmeas value after shelf life was
compared to the Fpred for the limits of the class according to the decay model. The classification was
considered: correct when the Fmeas value fell within the firmness interval predicted by the model for the specific
Mi class and acceptable when Fmeas values fell within the limits of the immediately adjacent M class (firmer,
Fmeas belonging to class Mi−1, softer, Fmeas belonging to class Mi+1); prediction related to Fmeas values which fell
within the Fpred limits for the Mi−2 class could be considered acceptable as fruit likely be ripen within a couple of
day, whereas the Fmeas values which fell outside the upper Fpred limit of the Mi−2 class (i.e. fruit which could
either ripen in a longer period or never ripen), and Fmeas values falling outside the lower Fpred limit of the Mi+1
class (fruit which would have a shorter shelf life than predicted and more prone to rot), make the prediction
unacceptable. Classification results for each class and model were expressed as percentage to total number
of fruit categorized in each class Mi.
3. Results and Discussion
3.1 Distribution of µa and classes of usability
The distribution of µa measured at harvest for the ‘Morsiani 90’ nectarines (Figure 1A) highlighted that the
range of µa is much smaller than that found in the early maturing cultivar ‘Spring Bright’ (Tijskens et al., 2006).
The value of µa of ‘Morsiani 90’ cultivar was low already at harvest, indicating a low chlorophyll content in the
pulp. Instead, firmness values decreased from 52±12 N at harvest to values around 10±5 N at the end of shelf
life similarly to what found in early maturing cultivars (Eccher Zerbini et al., 2006). Comparing the BSF for
firmness computed from parameters of the models for firmness decay prediction reported in Table 2 with that
of the early maturing ‘Spring Bright’ nectarine described by Rizzolo et al. (2009), all plotted in function of BSF
for µa (Figure 1B), it is evident the different synchronization of µa decrease and firmness decay among
‘Morsiani 90’ and ’Spring Bright’ cultivar, already at harvest, with a lower softening rate for the former cultivar
respect to that of the latter. For these reasons in order to study the classification performance of ‘Morsiani 90’
models it was not possible to use the µa limits of the usability classes established for ‘Spring Bright’ cultivar
(Rizzolo et al., 2009).
The softening trends of ‘Morsiani 90’ fruit during ripening after harvest at different µa intervals at the time of
harvest were considered (Figure 2). For ‘Morsiani 90’, being fruit softening slower than in ‘Spring Bright’, the
firmness values after 100 h as well as those at 198 h were considered in order to establish the usability
classes. Only few fruit exhibited µa>0.120 cm−1, and after 100 h firmness was still above 40 N; fruit having
µa<0.064 cm−1 were characterized by F<20 N at t=100 h and lower than 10 N at the last time, with the
µa<0.049 cm−1 ones showing F<30 N already at t=55h. Fruit with µa in the range 0.080-0.119 cm−1 after 100 h
had F values in the 20-50 N range, with the µa<0.089 cm−1 ones showing some fruit with F≈20 N already after
74 h, whereas fruit with µa in the range 0.065-0.079 cm−1 at 100 h had F<30 N. Crisosto et al. (2004) reported
that a firmness value of about 35 N is proper of a nectarine still firm enough to be transported home and ready
to buy, while values below 13.2 N indicate fruit ripe and soft (Crisosto et al., 2006). Basing on these
9
differences in firmness decay, the limits for the six usability classes from “transportable-hard” to “ready-to-eat-
soft” were established (Table 3).
(A) (B)
Figure 1: (A) Percentage distribution of absorption coefficient at 670 nm measured at harvest (480 fruits). (B)
Plot of BSF for firmness in function of BSF for µa for ‘Morsiani 90’ in comparison to ‘Spring Bright’ cultivar (SB,
Rizzolo et al., 2009)
Figure 2: Fruit firmness as a function of time at 20 °C for different levels of μa measured at harvest
Table 3: Lower limit of µa for each class Mi of predicted firmness potential for handling and eating, total
number of fruit and their percent distribution among classes for ‘Morsiani 90’ nectarines during shelf life at
harvest, and after storage at 0 °C and 4 °C
n Mi class Code
Lower µa
limit (cm−1)
Harvest
(% Nfruit)
0 °C storage
(% Nfruit)
4 °C storage
(% Nfruit)
1 transportable-hard TH 0.120 6.1 4.0 2.8
2 transportable T 0.090 18.4 20.1 20.7
3 ready-to-eat-firm-transportable RFT 0.080 27.4 28.9 28.3
4 ready-to-eat-firm RF 0.065 21.8 18.8 19.3
5 ready-to-eat-soft RS 0.050 22.9 24.8 25.5
6 ready-to-eat-very soft’ ORS <0.049 3.0 3.0 3.4
Nfruit 179 149 145
3.2 Prediction ability of models
Considering the results of regression analyses (Table 4), SEE ranged from about 4 % for the T 4C model to
about 7 % for the models T 0C and H+24h applied to fruit after storage at 0 °C. Comparing the results of the
three ripening, models after harvest and after storage at 4 °C had R2adj ranging from about 77 % for model T
harvest to about 65 % for model H+24h applied to fruit after storage at 4 °C. Moreover, MAE, AD, RMSED and
0.0
5.0
10.0
15.0
20.0
25.0
0
.1
7
0
.1
6
0
.1
5
0
.1
4
0
.1
3
0
.1
2
0
.1
1
0
.1
0
0
.0
9
0
.0
8
0
.0
7
0
.0
6
0
.0
5
0
.0
4
0
.0
3
%
f
ru
it
µa (cm
−1)
-20
-15
-10
-5
0
5
10
-6 -4 -2 0 2 4
∆
t*
F
∆t*μa
∆t*FSB
∆t*F mod T harvest
∆t*F mod H+24h
∆t*F mod T 0°C
∆t*F mod T 4°C
10
s/RMSED values indicate that models T harvest and T 4C had higher performance than H+24h and also that
H+24h model for ripening at harvest had higher performance than for ripening after storage. In contrast,
results of linear regression for models after storage at 0 °C indicate a lower performance of both the T 0C and
H+24h models respect to the other ripening. These results are in agreement with the findings on these models
discussed by Eccher Zerbini et al. (2011).
Table 4: Results of linear regression models between Fmeas and Fpred from µa at harvest for fruit ripened after
harvest and after storage at 0 °C and 4 °C
Ripening After harvest After storage at 0 °C After storage at 4 °C
Model T harvest H+24h T 0C H+24h T 4C H+24h
Intercept estimate 3.84 0.75 13.07 4.56 4.52 7.56
SE (sign) (0.901)*** (0.956)ns (1.391)*** (1.462)** (0.20)*** (1.089)***
Slope estimate 0.63 0.66 0.59 0.59 0.72 0.94
SE (sign) (0.026)*** (0.027)*** (0.040)*** (0.042)*** (0.038)*** (0.057)***
R2adj 77.32 77.00 58.89 57.03 71.65 65.24
SEE 6.349 6.732 7.207 7.573 4.428 6.704
MAE 4.862 5.080 5.867 5.987 3.106 4.924
AD 29.11 34.15 34.86 33.42 27.98 61.03
RMSDE 11.7571 13.0612 9.4001 12.6510 5.1689 9.3911
s/RMSDE 1.58 1.42 1.57 1.17 1.89 1.04
Figure 3: Comparison of classification results for ripening (SL) at harvest and after storage at 0 °C and 4 °C
obtained with firmness decay models of Table 2
Figure 3 shows the results of the test for misclassification for the three ripening. Considering the classification
results for ripening at harvest, model T harvest classified acceptably (i.e. in the range Mi–2 – Mi+1) more than
70 % of fruit of ORS class, more than 87 % of fruit of RFT and T classes, and all the fruit of TH class, whereas
model H+24h at harvest classified acceptably the totality of fruit of RFT, T and TH classes, but only about
64 % of ORS fruit. The Fmeas values of misclassified fruit of ORS, RS and RF classes corresponded to Fpred
values for RFT, T and TH classes, respectively. After storage at 0 °C, model T 0C correctly predicted the fruit
of ORS class, and acceptably classified more than 85 % of fruit of RS, RFT and TH classes; for the
misclassified fruit of RF and T classes the Fmeas corresponded to Fpred for T and ORS classes, respectively. If
model H+24h is applied to predict softening of fruit after storage at 0 °C, more than 90 % of fruit of RFT and T
S
L
H
A
R
V
E
S
T
S
L
a
ft
e
r
s
to
ra
g
e
0
°C
S
L
a
ft
e
r
s
to
ra
g
e
4
°C
0% 20% 40% 60% 80% 100%
ORS
RS
RF
RFT
T
TH
model T harvest
0% 20% 40% 60% 80% 100%
ORS
RS
RF
RFT
T
TH
model H+24h
0% 20% 40% 60% 80% 100%
ORS
RS
RF
RFT
T
TH
model T 0C
0% 20% 40% 60% 80% 100%
ORS
RS
RF
RFT
T
TH
model H+24h
0% 20% 40% 60% 80% 100%
ORS
RS
RF
RFT
T
TH
model T 4C
0% 20% 40% 60% 80% 100%
ORS
RS
RF
RFT
T
TH
model H+24h
>Mi-2 Mi-2 Mi-1 Mi Mi+1 Mi+2 >Mi+2
11
classes and the totality of fruit of TH class were acceptably predicted, whereas 33 % of ORS fruit had Fmeas
corresponding to Fpred for RF class, 46 % of RS fruit showed Fmeas corresponding to Fpred for T class and 33 %
of RF nectarines had Fmeas corresponding to Fpred for TH class. The models T 4C and H+24h applied to predict
softening after storage at 4 °C showed the worst performance in classifying nectarines: model T 4C correctly
predicted the totality of ORS fruit, 20 % of TH fruit showed Fmeas values corresponding to Fpred for T class and
fruit of the other classes were misclassified in proportions ranging from about 33 % of RS class to about 50 %
of RF class. If model H+24h is applied to predict softening of fruit after storage at 4 °C, only the totality of ORS
fruit and more than 90 % of RS nectarines were acceptably predicted, whereas the majority of fruit of T and
TH classes showed Fmeas values corresponding to Fpred for RF and RS classes, respectively. The lower
performance of models T 4C and H+24h for fruit after storage at 4 °C respect to those of the other two
ripening could be due to the fact that in these fruit chilling injury symptoms appeared (Lurie et al., 2011), and
that the same changes in cell wall metabolism which induce the appearance of chilling injury also affect
firmness and softening rate (Eccher Zerbini et al., 2011), so having a negative impact on the prediction ability
of the kinetic model.
4. Conclusions
Results suggest that the methodology based on the μa measured by TRS at harvest and its conversion into
BSF might be used as a management tool in the supply chain of late-maturing nectarines. The model H+24h
showed a classification ability very close to that of T models based on data of the whole shelf life, and was
able to correctly segregate the RFT, T and TH classes for ripening at harvest and after storage at 0 °C, and
the ORS and RS classes for ripening after storage at 4 °C.
References
Crisosto C.H., Garner D., Andris H.L., Day K.R., 2004, Controlled delayed cooling extends peach market life,
HortTechnol. 14, 99-104
Crisosto C.H., Crisosto G., Echeverria G., Puyc J., 2006, Segregation of peach and nrctarine (Prunus persica
(L) Batsch) cultivars according to their organoleptic characteristics, Postharvest Biol. Technol. 39, 10-18
Crisosto C.H., Valero D., 2008, Harvesting and postharvest handling of peaches for the fresh market, The
Peach: Botany, Production and Uses, Eds. Layne D.R., Bassi D., CABI, Wallingford, UK
Cubeddu R., D’Andrea C., Pifferi A., Taroni P., Torricelli A., Valentini G., Dover C., Johnson D., Ruiz-Altisent
M., Valero C., 2001, Non-destructive quantification of chemical and physical properties of fruits by time-
resolved reflectance spectroscopy in the wavelength range 650-1000 nm, Appl. Opt. 40, 538-543
Eccher Zerbini P., Vanoli M., Grassi M., Rizzolo A., Fibiani M., Cubeddu R., Pifferi A., Spinelli L., Torricelli A.,
2006, A model for the softening of nectarines based on sorting fruit at harvest by time-resolved reflectance
spectroscopy, Postharvest Biol. Technol. 39, 223-232
Eccher Zerbini P., Vanoli M., Rizzolo A., Jacob S., Torricelli A., Spinelli L., Schouten R.E., 2009, Time-
resolved reflectance spectroscopy as a management tool in the fruit supply chain: an export trial with
nectarines, Biosystems Engineering 102, 360-363
Eccher Zerbini P., Vanoli M., Lovati F., Spinelli L., Torricelli A., Rizzolo A, Lurie S., 2011, Maturity assessment
at harvest and prediction of softening in a late maturing nectarine cultivar after cold storage. Postharvest
Biol. Technol. 62, 275-281
Lurie S., Vanoli M., Dagar A., Weskler A., Lovati F., Eccher Zerbini P., Spinelli L., Torricelli A., Feng R.,
Rizzolo A., 2011, Chilling injury in stored nectarines and its detection by time-resolved reflectance
spectroscopy, Postharvest Biol. Technol. 59, 211.218
Rizzolo A., Vanoli M., Eccher Zerbini P., Jacob S., Torricelli A., Spinelli L., Schouten R.E., Tijskens L.M.M.,
2009, Prediction ability of firmness decay models of nectarines based on the biological shift factor
measured by time-resolved reflectance spectroscopy, Postharvest Biol. Technol. 54, 131-140
Tijskens L.M.M., Heuvelink E., Schouten R.E., Lana M.M., van Kooten O., 2005, The biological shift factor.
Biological age as a tool for modeling in pre- and postharvest horticulture, Acta Hort. 687, 39-46
Tijskens L.M.M., Eccher Zerbini P., Vanoli M., Jacob S., Grassi M., Cubeddu R., Spinelli L., Torricelli A., 2006,
Effects of maturity on chlorophyll related absorption in nectarines, measured by non-destructive time-
resolved reflectance spectroscopy, Int. J. Postharvest Technol. Innov. 1, 178-188
Tijskens L.M.M., Eccher Zerbini P., Schouten R.E., Vanoli M., Jacob S., Grassi M., Cubeddu R., Spinelli L.,
Torricelli A., 2007. Assessing harvest maturity in nectarines, Postharvest Biol. Technol. 45, 204-213
Torricelli A., Spinelli L., Contini D., Vanoli M., Rizzolo A., Eccher Zerbini P., 2008, Time-resolved reflectance
spectroscopy for non-destructive assessment of food quality, Sens. Instr. Food Qual. Safety 2, 82-89
12