You are on page 1of 9

Abstract (0.76 ≥ R2≥ 0.

58), and chlorophyll

content in green cultivars (0.88 ≥ R2≥
The pigment content of mango peel is 0.62).Results showed that reflectance
an important quality determinant due to measurements could be used to predict
its involvement in the nutritional value mango quality determinants such as the
and coloration of the fruit. In addition, fruit maturity stage according to the
pigment content in the peel is related to peel chlorophyll content, the fruit
the fruit maturity stage and has been coloration and the content of valuable
suggested as a reliable harvest components, i.e., anthocyanins and
criterion. Measurements of pigment carotenoids.
content by vis-spectrophotometry and
chromatography methods are
destructive, time-consuming and
expensive. This study deals with the 1. Introduction
potential of reflectance measurements The chlorophyll content in mango peel
as a non-destructive method to assess was recently found to be a reliable
pigment content and color in mango indicator of the mango maturity stage
fruit peel. We also investigated the (Léchaudel et al., 2010). In addition,
relationship between the pigment anthocyanins and carotenoid pigments
content of the peel and its color. Eight of mango fruit have been the object of
mango cultivars, i.e., Cogshall, Kent, increasing interest since they are
Caro, Sensation, Tommy Atkins, Nam known to provide health benefits due to
Doc Mai, Irwin and Heidi, were studied. their antioxidant activity (Ajila et al.,
Hue angle distribution was used to 2007a,b, 2010; Berardini et al.,
characterize and to compare cultivar 2005a,b).Spectrophotometric and
colors. The peel contents in HPLC measurements were used to
anthocyanin, carotenoids, chlorophyll a measure pigment content in mango
and chlorophyll b were predicted using
peel and revealed their variations
Partial Least Square Regressions between cultivars (Ajila et al., 2007a;
based on reflectance indexes and Berardini et al., 2005a,b;Ketsa et al.,
compared between cultivars. Our 1999) and fruit maturity stages (Ketsa
results showed that reflectance et al., 1999;Medlicott et al., 1986).
measurements can accurately (R2≥ However, these methods are
0.91, RMSE ≤ 5.74 _g gFW−1) predict destructive, time-consuming and
the contents of the main pigment expensive. Non-destructive
contained in the mango peel. Further measurements of surface reflectance in
studies are nevertheless required to
the visible range have been suggested
increase PLSR robustness. Significant to predict the pigment content of leaves
differences were found in the hue angle and apple skin (Merzlyak et al.,2003a;
distribution and the predicted peel Mielke et al., 2012; Gitelson et al.,
pigment content between the eight 2001; Sims and Gamon,2002; Zude,
mango cultivars studied. All 2003). Such measurements could also
relationships between pigments and be used to assess the content of the
hue angle values were found to be main pigments of mango fruit peel, i.e.,
cultivar-dependent. The highest chlorophyll a, chlorophyll b, carotenoids
correlations between hue angle values and anthocyanins (Medlicottet al.,
and pigment contents were found for 1986). These pigments are involved in
anthocyanin content in red cultivars the photosynthesis activity, in the
photo-protection of the fruit (Merzlyak et production season in the CIRAD
al., 2002;Steyn et al., 2002) and in the orchard collection of Reunion Island
fruit coloration (Lancaster et al., (20°52’48’’S, 55°31’48’’E), consisting of
1997).Fruit color is an important factor 7-year-old trees grafted on ‘Maison
for market acceptance (Nguyenet al., Rouge’ cultivars. Trees were well-
2004; Vásquez-Caicedo et al., 2002; irrigated, spaced 5 m × 6 m, and
Litz, 2009) and could be used as a tool approximately 3 m high. Fruits were
to describe mango cultivars (Ayala- randomly chosen on the tree in order to
Silva et al.,2005) or to predict mango represent the variability of sun
maturity stages (Jha et al., 2007). exposure conditions, and harvested at
Colorimetric coordinates can be the optimal harvest stage for the local
deduced from reflectance market, considering that fruits are at
measurements using color-matching their prime during the next 3–5 days
functions (Vos, 1978). These functions (Léchaudel and Joas, 2006).Just after
make it possible to calculate the the harvest, approximately four NIR
quantity of red, green and blue colors of spectra of fruit peel surface at random
a reflectance spectrum in order to locations on each fruit were monitored
reproduce the sensitivity of the human from350 to 2500 nm with the contact
eye (Abbott, 1999). Several authors probe of a portable spectrometer
found relationships between (LABSPEC 2500, Analytical Spectral
colorimetric measurements and Devices, Inc., Boulder, CO,USA). For
pigment content in the skin (Lancaster 33 supplementary samples
et al., 1997; Iglesias et al., 2012; Steyn representing the observed color
et al., 2004) or flesh tissues (Ornelas- variability observed within the studied
Paz et al., 2008; Vásquez-Caicedo et cultivars, pigment concentrations in the
al.,2005; Sánchez et al., 2006) for peel were determined after extraction
various fruits and vegetables. by visspectrophotometry in order to
develop predictive models based on
In this paper, reflectance reflectance spectra. NIR spectra were
measurements were used to evaluate monitored on these skin samples prior
the relationships between the color and to their destructive analysis to
the pigment content of mango peel. The determine pigment content.
study is based on eight mango cultivars
in order to ensure a wide range of XYZ colorimetric coordinates of skin
variation in fruit colors and pigment color were obtained from its reflectance
contents. spectra in the visible range
(Colorimetry, 2004), where S(δ) is the
2. Materials and methods spectral reflectance of the measured
2.1. Fruits samples, measurements of mango peel, I(δ) is the spectral
reflectance spectra and color irradiance of the illuminant, and x(δ),
determination y(δ) and z(δ)are the color-matching
functions. The CIE standard illuminant,
The study was carried out on 61, 60, D65, was used in this paper (Xn =
79, 60, 51, 51, 51, and 58 mangoes 0.9503, Yn = 1 and Zn = 1.0891). The
from cv. Cogshall, Kent, Caro, calculated XYZ color coordinates were
Sensation, Tommy Atkins, Nam Doc then converted to CIELAB coordinates,
Mai, Irwin and Heidi, respectively. Fruits and the hue angle value (H◦) was
were grown during the 2012–2013
calculated from a*and the b*
coordinates using Eq. (2).

Anthocyanin pigments were extracted

according to the procedure described
by Jing and Giusti (2007). Briefly,
approximately0.5 g of crushed mango
peel was added to 4 ml of deionized
water. The sample was shaken in a dry
bath at 50°C for 1 h.

The resulting extract was filtered

through a Whatman No.1 filter paper
under a vacuum using a Büchner
2.2. Determination of peel pigment funnel. Extraction with deionized water
concentration provides the highest yield in
comparison to other extraction solvents
The photosynthetic pigments, (Jing and Giusti, 2007).
chlorophylls a and b, as well as the total
carotenoids (x + c), were extracted with The monomeric anthocyanin content of
100% dimethylsulfoxide (DMSO). the mango peel was measured using a
Approximately 5 cm 2 of mango peel spectrophotometric pH differential
samples were removed with a peeler on protocol (Leeet al., 2005). A VIS
each fruit studied, and then immediately scanning spectrophotometer (Helios
frozen in liquid nitrogen and stored in Epsilon, Thermo Fisher Scientific) was
an ultra-freezer (−80°C) until chlorophyll used to measure absorbance at 520
and carotenoid extraction and and 700 nm against a distilled water
quantification were performed. blank. Anthocyanin content was
Approximately 0.5 g of the mango peel expressed as micrograms of cyanidin-
samples were extracted with 5 ml of 3-glucoside equivalent per gram of
DMSO after incubation in the dark, in a fresh mango peel.
dry bath at 65°C for 5 h. Heated
samples were removed from the dry 2.3. Prediction of peel pigment
bath and allowed to reach room content using reflectance spectra
temperature. The absorbance (A)
In order to non-destructively predict the
ofsupernatants was then measured at
peel contents in pigments using
480, 649, and 665 nm with a VIS
reflectance spectra, a model based on
scanning spectrophotometer (Helios
Partial Least Square Regression (PLS)
Epsilon, Thermo Fisher Scientific,
was established. The tested
USA). To calculate the chlorophyll a
preprocessing methods (first and
([Ca]), chlorophyll b ([Cb]) and total
second derivations) did not improve the
carotenoid ([Cx+c]) content, the
models accuracy so no transformed
equations of Wellburn(1994) were used
spectral data (log(1/reflectance)) were
(Eq. (3)).
used to build models.
Model calibrations were performed from
the destructive measurements of
pigment content of the 33 epidermis
samples. As proposed by several
authors (Merzlyak et al., 2003a; Mielke
et al.,2012; Gitelson et al., 2001; Sims
and Gamon, 2002), the PLS regression
was calculated from reflectance
indexes based on optical pigment
features. The number of latent variables
for the PLSR model was obtained by
using leave-one-out cross-validation to
avoid over-fitting of the equation. The
procedure provided by the PLS
package (Mevik and Wehrens, 2007)
developed in R soft-ware (Team, 2012)
and fully described by Cornillon (2010)
was applied. Model prediction error was
evaluated using the Root Mean Square
Error (RMSE) and the coefficient of
determination (R2) as indicators.
Calculation of the RMSE is described in
Eq. (4), where yt is the t-th observed or
reference value,y^t is the t-th simulated
value, and n is the number of observed
or simulated values.
Fig 1.

Peel reflectance average in the visible range(solid line) ± standard deviation(dashed lines) calculated from all mango
cultivars studied (A) completed by diagram of the color of the visible spectra, and by a diagram of the maximal
absorbance of pigments(B), i.e., chlorophyll a (431.5 and 665.6 nm),chlorophyll b(460.3 and 647.6 nm)(wellburn,1994),
anthocyanins (550 nm) (merzlyak et al, 2003b), and carotenoids(479, 456 and 449nm) (wellburn,1994).(C-J) peel
reflectance measurements in the visible range according to the mango cultivars studied. The color of each curve
corresponds to the color of the peel sample. (For interpretation of the references to colors in this figure legend, the
reader is referred to the web version of this article.)

Models calibrated using the 33 peel hue angle values were compared
samples were applied to all reflectance between cultivars according to the non-
measurements in order to predict the parametric test of Kolmogorov–Smirnov.
peel contents in pigments of cultivars
studied. Multiple comparisons between cultivars
of the predicted pigment content
2.4. Statistical analysis averages were performed using Tukey’s
test performed with the HSD test R
All analyses and graphical renderings function provided in the Agricolae
were performed with R software (Team, package. For all statistical analysis, the
2012) implemented with Colorspace null hypothesis was rejected, i.e.,
(Ihaka et al.,2008) and PLS packages absence of effect or difference, when the
(Mevik and Wehrens, 2007). The P-value was lesser or equal to the
probability density function of the hue significant level of 0.05.
angle values was calculated for each
cultivar using the density function of R
software. The probability densities of the
3. Results and discussion

3.1. Analysis of reflectance spectra

The average spectra of peel cultivars in

the visible region(Fig. 1A) had a pattern
similar to those found in apple
fruits(Merzlyak et al., 2003b) and leaves
(Gitelson et al., 2002). The reflectance
was more variable and higher in the red
region(600–800 nm) and in the green
region (500–600 nm) than in the blue
region (400–500 nm) (Fig. 1A). In fact,
reflectance was found to fluctuate for all
cultivars between 0.04 and 0.26 at 450
nm, between 0.04 and 0.63 at 550 nm,
and between 0.06 and 0.77 at 673 nm
(Fig. 1C–J). In the visible range, the
reflectance of fruit peels fluctuates with
the variation in the pigment contents
(Merzlyaket al., 2003b). The increase of
the reflectance in the red region(640–700
nm) can be mainly linked to a decrease in
the chlorophyll content (Fig. 1A) that
absorbs radiations in this region
(Merzlyaket al., 2003b) (Fig. 1B).
However, since the chlorophyll pigment
also absorbs radiations near 450 nm (Fig.
1B), it was expected that the presumed
chlorophyll variation suggested by the
variation in the
Fig. 2. Hue angle probability function (colored curve) and histogram (bars) calculated for each cultivar. Values indicated
next to the curve correspond to the hue angle value of the peaks. (For interpretation of the references to color in this

figure legend, the reader is referred to the web version of this article.)

red region would also be observed in the pigment contents, as previously

blue region (Fig. 1A and B). Lack of a discussed. Color and reflectance
relationship between the variation in the spectrum variations were found to be
blue and the red regions may be related in some cases. For example,
explained by the presence of carotenoid yellow coloration was associated with a
pigments that absorb radiations in the higher reflectance in the red region than
blue region, like chlorophyll (Wellburn, in the green one for Caro mangoes (Fig.
1994; Gitelson et al., 2002) (Fig. 1B). 1H), suggesting a lower chlorophyll
Merzlyak and Solovchenko (2002) content. Furthermore, red curves were
suggested that the carotenoid synthesisis associated with a lower reflectance near
induced when the chlorophyll degradation 550 nm than green ones (Fig. 1C, D, F, G
occurs during fruit senescence in order to and J), which can be linked to higher
provide protection to light-sensitive anthocyanin content. However, it can be
constituents of plant tissues from observed for all cultivars that the same
radiation in the blue part of the visible color can result from variations in spectral
spectrum. distribution, especially for Tommy
mangoes (Fig. 1G), which exhibited a
The reflectance variation near 550 nm quite homogenous red coloration despite
(Fig. 1A)can be linked to the variation in various reflectance curves in the red
anthocyanin absorption (Gitelsonet al., region.
2001; Merzlyak et al., 2003b) (Fig. 1B).
3.2. Peel color variations
It can be observed in Fig. 1C–J that
reflectance spectra varied between A wide color variation was found between
mango cultivars, suggesting different cultivars and within cultivars (Fig. 2).
The commercial success of mango associated with each average value.
cultivars depends partly on their external Comparisons of the peel colors based on
coloration as most cultivars intended for the distribution of the hue angle values
the Southeast Asia market have green (Fig. 2) made it possible to significantly
skins, whereas the ones intended for the differentiate each cultivar from the others.
northern market have a red coloration
(Litz, 2009). These results suggested that the study of
the hue angle distribution is an effective
Among the eight cultivars studied, only method for characterizing mango cultivar
Caro (Fig. 2A) and Nam Doc Mai (Fig. color since it takes their color
2B) did not show a red coloration, i.e., heterogeneity into account (Fig. 2).
hue angle ≤50°.
3.3. Measurements and non-destructive
Hue angle distribution differed prediction of peel pigment content
considerably between cultivars (Fig. 2)
with regard to their range of variations Use of reflectance measurements to
and pat-terns. For example, hue angle access pigment contents requires a
values ranged from 60°(yellow skin)to calibration phase, but has the advantage
126◦(green skin) for Nam Doc Mai of being non-destructive and convenient.
For 33 samples, pigment concentrations
mangoes (Fig. 2B), compared to −30◦
in the peel were determined after
(purple skin) to 116◦(green skin) for extraction by vis-spectrophotometry in
Sensation mangoes(Fig. 2H). The order to develop predictive models based
proportion of red coloration varied on reflectance spectra. Measurements
between cultivars(Fig. 2) and sorting varied from 1.97 to 99.29 _g gFW−1for
cultivars according to their proportion of
chlorophyll a, and from 1.81 to 58.7 _g
blush area (H◦< 50◦) in decreasing order
gFW−1 for chlorophyll b, in accordance
resulted in the following ranking: Tommy
with Wang et al. (2008) who found that
Atkins, Irwin, Sensation, Heidi, Kent,
total chlorophyll contents fluctuated
Cogshall, Nam Doc Mai and Caro, with
between 33.8 and 199.1 _g gFW−1 in
85.5%, 77.6%, 55.1%, 51.1%, 44.3%,
the peel of Tainong mangoes. Carotenoid
33.7%, 0%and 0% of blush area,
respectively. Multiple comparisons
(Tukey’stest, P < 0.05) of the hue angle
averages of cultivars indicated sig-
nificant differences. However, no
distinction was made between Caro and
Nam Doc Mai, Cogshall and Kent, Heidi
and Sensation, and Irwin and Tommy
Atkins (data not shown). Ayala-Silva et
al.(2005) have already observed the limit
of the use of the hue angle averages for
comparing colors between cultivars
because few significant differences were
found due to the large standard errors
Table 1 Reflectance indexes used to predict chlorophyll a, chlorophyll b, carotenoid, and anthocyanin
content in mango peel, and NIR calibrations performances, with R2, the determination coefficient, RMSE,
the Root Mean Square Error and n the number of samples.

content in the fresh peel varied from different pig-ments in the visible range (Table 1). Various
indexes are proposedin the literature. The ones used in this
14.37 to 35.9 _g gFW−1, in line with the study were inspired bythose found by Merzlyak et al.
measurements of Ketsa et al. (1999) who (2003a,b). Since reflectance dataare composed of many noisy
found that the β-carotenoid content, i.e., and collinear variables, the use of PLSregression is
recommended (Wold et al., 2001) and commonly used(Nicolaï
the main carotenoid pigment in mango et al., 2007). Moreover, multivariate analysis, such as
peel (Medlicott et al., 1986), for Nam Doc PLSregression, of absorbance values was found to improve the
Mai and Tongdum mangoes varied from pre-diction of pigment performed with linear regression
(Zude-Sasseet al., 2002). The cross validation leave-one-out
5 to 25 _g gFW−1 of fresh peel. method used todetermine the number of latent variables for
Concerning anthocyanin content in the the PLSR (Fig. 3)made it possible to maximize the predictive
peel, measurements varied from 0 ability of the modeland to avoid model over-fitting (Cornillon,
2010). The averageRMSE ratios of measured pigment contents
to100.6 _g gFW−1. Ajila et al. (2007a) were 6.8%, 5.9%, 10.3%,and 15.3%, for chlorophyll a,
reported anthocyanin contents in the dry chlorophyll b, carotenoid and antho-cyanin, respectively.
mango peel varying between 2030 and Higher prediction errors were observed foranthocyanin
pigment (Fig. 3D) compared to the other pigments(Fig. 3A–C).
5650 _g gDW−1,contrary to Berardini et The assessment of pigment contents using peel
al. (2005a,b) who reported anthocyanin spectralmeasurements assumes that the optical pigment
contents varying between 0.211 and 3.72 features remainthe same in the different cultivars and
through different stages offruit development (Merzlyak et al.,
_g gDW−1 for several mango cultivars, 2003b). However, anthocyaninoptical features are affected by
and to Dorta et al. (2012) who reported several factors such as pH, tem-perature, light and presence
of enzymes, flavonoids and metallicions (Casta˜neda-Ovando
values fluctuating between 15 and 33 _g
et al., 2009). These variations can explainFig
gDW−1 for ripe Keitt mangoes
.Differences between authors are
probably related to the differences in
extraction and measurement methods
Several authors (Mielke et al., 2012; Gitelson et al.,
2001;Merzlyak et al., 2003b) estimated the pigment content
in the peelusing indexes related to the optical features of the

Related Interests