- Email: [email protected]
Segregation of apricots for storage potential using non-destructive technologies
Postharvest Biology and Technology 86 (2013) 17–22
Contents lists available at ScienceDirect
Postharvest Biology and Technology journal homepage: www.elsevier.com/locate/postharvbio
Segregation of apricots for storage potential using non-destructive technologies Jinquan Feng a,∗ , Jill Stanley b , Mohammed Othman c , Allan Woolf a , Maureen Kosasih d , Shane Olsson a , Graeme Clare a , Nick Cooper c , Xirui Wang e a
The New Zealand Institute for Plant & Food Research Limited, Mt Albert Research Centre, Private Bag 92169, Auckland 1142, New Zealand The New Zealand Institute for Plant & Food Research Limited, Clyde Research Centre, 990 Earnscleugh Road, RD 1, Alexandra 9391, New Zealand c Taste Technologies Ltd., 64-72 Victoria Street, Onehunga, Auckland 1061, New Zealand d Westland Milk Products, 56 Livingstone Street, Hokitika 7810, New Zealand e Rural Science and Technology Development Centre of Shaanxi Province, Xian, China b
a r t i c l e
i n f o
Article history: Received 16 March 2012 Accepted 13 June 2013 Keywords: Storage life Flesh firmness Visible-near infrared spectroscopy Acoustic firmness Impact firmness Fruit colour
a b s t r a c t This study was set up to identify critical maturity indices affecting storage potential of apricots and demonstrate the potential for using non-destructive measurements to segregate harvested crops for sequential marketing. Fruit of two apricot (Prunus armeniaca) cultivars (‘Clutha Gold’ and ‘Genevieve’) were harvested and stored for four weeks at 0 ◦ C followed by four days of simulated shelf life at 20 ◦ C. Fruit colour, acoustic firmness, impact firmness, flesh firmness (FF0 ), dry matter content and soluble solids content measured non-destructively at harvest were correlated to the flesh firmness measured at the end of refrigerated storage and simulated shelf life (FFFinal ) through stepwise regression. The regression models indicated that FF0 is a predominant factor determining FFFinal . According to the exponential model describing the relationship between FF0 and FFFinal , ‘Genevieve’ and ‘Clutha Gold’ could be stored at 0 ◦ C for four weeks if harvested at firmness above 47 or 56 N, respectively. Segregation of harvested crops according to FF0 estimated from VNIR would enable sequential marketing of fruit according to storage potential to reduce fruit loss. © 2013 Elsevier B.V. All rights reserved.
1. Introduction Apricot (Prunus armeniaca L.) is one of the most popular summer fruit grown in temperate climate zone with a total world production of about 3.8 million tonnes (FAO 2009). New Zealand currently exports 1000–1500 tonnes of fresh apricots annually worth $NZ7-9 million, which accounts for about 2% of world trade in fresh apricots. Apricots are climacteric fruit with limited storage life of 1–4 weeks (Fan et al., 2000). Mixed maturity at harvest affects storage life and fruit quality in the market, which causes concern (Bruhn et al., 1991; Fan et al., 2000; Infante et al., 2008). Fruit background colour, soluble solids content (SSC) and flesh firmness (FF) have been considered important maturity indices (Visagie, 1988; Brown and Walker, 1990; Biondi et al., 1991; Aubert and Chanforan, 2007; Infante et al., 2008; Gouble et al., 2010). However, the relative importance of these maturity indices for storage potential has not been established. Traditionally, apricots were harvested and graded based on visual assessment of fruit colour. For example, ‘Palsteyn’ apricots
∗ Corresponding author. Tel.: +64 9 9258618; fax: +64 9 9257001. E-mail address: [email protected] (J. Feng). 0925-5214/$ – see front matter © 2013 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.postharvbio.2013.06.015
with orange-yellow skin colour were considered ripe and providing the highest acceptability, and those harvested at an intermediate stage of maturity (light yellow skin colour) provided an average acceptability, while unripe fruit (greenish skin colour) were not acceptable (Infante et al., 2008). However, visual assessment of fruit colour is subjective and can be affected by environmental factors such as sunlight and temperature. Fruit colour is not a reliable representation of flesh firmness or soluble solid content and therefore is only an approximate indication of maturity at harvest. For long-term storage or transport, fruit should be harvested at the pre-climacteric stage, before they attain their full flavour and colour, and are more tolerant to handling and prolonged cold storage (Aubert and Chanforan, 2007). Therefore, the optimal maturity at harvest for each cultivar should be a balance between eating quality and storage potential. Identifying key factors affecting storage potential, and developing a quantitative model to describe the relationship between maturity indices at harvest and fruit quality at the end of targeted storage duration, are important steps to optimising harvest maturity. Recent developments in non-destructive technologies to measure maturity indices offer the possibility of non-invasive on-line packinghouse screening of colour, SSC and firmness at harvest (Muramatsu et al., 1996; Carlini et al., 2000; McGlone and Jordan,
18
J. Feng et al. / Postharvest Biology and Technology 86 (2013) 17–22
2000; Camps and Christen, 2009; Bureau et al., 2010; Berardinelli et al., 2010). These same technologies provide the opportunity to measure both maturity indices at harvest and fruit quality after storage on the same fruit, and to investigate the relationship between these on an individual fruit basis. It was hypothesised that the relationship between maturity indices at harvest and fruit quality at the end of storage could be revealed more clearly on an individual fruit basis than on a fruit batch basis because of the large differences among individual fruit and the elimination of sample error. In this study, a total of 450 fruit from each cultivar were measured using three types of non-destructive measurement technologies at harvest. FF, SSC and dry matter content (DM) were measured destructively on a subsample of 180 fruit from each cultivar immediately after non-destructive measurements so that to develop calibration models to estimate these attributes for the remaining fruit (270 fruit per cultivar) that were stored for four weeks at 0 ◦ C followed by four days of simulated shelf life at 20 ◦ C. Fruit quality measured at the end of refrigerated storage and simulated shelf were correlated with maturity indices measured non-destructively at harvest on an individual fruit basis. 2. Materials and methods 2.1. Fruit harvest Four hundred and fifty apricot fruit for each of the two cultivars (‘Clutha Gold’ and ‘Genevieve’) were harvested from Central Otago, a major apricot production area in New Zealand. Fruit were graded into three maturity classes based on visual assessment of fruit colour (background colour and percentage of blush) and packed separately into single layer trays with plastic Plix® trays. This separation was aimed to reduce maturity variation within each package to minimise the effect of released ethylene from more mature fruit contributing to the softening of less mature fruit. It also ensured that a range of fruit maturities was represented in the sample. Packed fruit were sent to the Plant & Food Research in Auckland on the same day by airfreight. Fruit were kept at ambient temperature overnight before measurements were made.
Fruit colour was measured on opposite flat sides of each fruit along the equatorial zone using a Minolta Chroma Metre (CR300, Konica Minolta Sensing Americas, Inc., USA) equipped with 6-eliment silicon photocells (detector) operated with D65 light source and calibrated with a white calibration plate. Measurement on each fruit was taken on the blushed side, then the shaded side. Measured values were expressed in L (lightness), C (chroma) and H (◦ hue) colour space. The two readings from each fruit were averaged to give one reading for each fruit (i.e. L0 , C0 and H0 , where “0” indicates that the parameter was measured when storage time = 0 days. This applies to all the fruit attributes measured at harvest). Acoustic firmness and impact firmness were measured using an AWETA AFS (AWETATM Impact & Acoustic Firmness System, Nootdorp, Holland). The AFS is a small bench top unit that gently taps the fruit and ‘listens’ to the acoustic response. The vibration pattern (resonance attenuated vibration) of the fruit is analysed and translated into an acoustic firmness (AF0 , Eq. (1)) that is characteristic of the overall stiffness of the fruit (Cooke, 1972). The AFS also measures fruit weight and impact firmness (IPF0 , a measure of the local fruit surface elasticity). Two AFS measurements were made on two flat sides of each fruit along the equatorial plane. The two readings from each fruit were averaged to give a single value for each fruit. 2/3
2.3. Destructive measurements at harvest Destructive measurements of soluble solids content (SSC0 ), dry matter content (DM0 ) and flesh firmness (FF0 ) were made on 180 fruit per cultivar to provide data to develop calibration models to predict FF0 , SSC0 and DM0 for the other 270 fruit per cultivar that went to refrigerated storage and simulated shelf life test. FF0 was measured on two sides of each fruit along the equatorial zone using a 7.9-mm diameter probe attached to a GUSS FTA (GUSS Manufacturing Ltd., South Africa). Penetration speed, trigger force and penetration distance were set to 10 mm s−1 , 50 g and 8 mm, respectively. A patch of skin about 1 mm in thickness was removed from each side of the fruit before firmness measurement. SSC0 was measured using a Digital Hand-Held “Pocket” Refractometer (Model PAL, Atago, Tokyo, Japan) with combined juice expelled from FF0 measurements on each fruit. Dry matter content (DM0 ) was measured using a cross section slice (including skin and flesh, but excluding the stone) cut along the equator of each fruit. The slice was weighed immediately after cutting and then after drying to a constant weight at 65 ◦ C over a period of 24 h. 2.4. Storage and final firmness assessment
2.2. Non-destructive measurements at harvest
AF0 = Fo20 M0
where Fo0 is the resonant frequency of the first elliptical mode (Hz) and M0 is the mass of the fruit (kg). Visible-near-infrared (VNIR) spectra measurements were carried out using a commercial VNIR grading system (Taste Technologies, New Zealand). A continuous beam of light generated from a halogen lamp was focused on fruit passing through the VNIR unit while spectra in a wavelength range of 300–1100 nm with a resolution of 10 nm were taken in a reflectance mode. Each fruit passed through the VNIR unit twice and two VNIR spectra were taken on each fruit. Each plectrum was normalised to reference spectra taken when there were no fruit loaded and transformed into a second derivative spectrum (Savitsky and Golay, 1964; Steinier et al., 1972) before calibration and prediction. These are the most commonly used data transformations necessary to achieve robust calibration models.
(1)
The remaining 270 fruit per cultivar were stored at 0 ◦ C for four weeks and subjected to four days of simulated shelf life at 20 ◦ C under ambient light before the final assessment of flesh firmness (FFFinal ) and disorders such as rots, soft patches and chilling injury that render fruit unsalable. 2.5. Data analysis Calibration models to predict FF0 , SSC0 and DM0 from VNIR spectra were developed based on data from 130 fruit per cultivar. Data from the other 50 fruit per cultivar destructively measured at harvest were used as cross validation data. Calibration models were established using partial least square (PLS) regression between second derivative spectra and the reference data (SSC0 , DM0 and FF0 ). The reference data measured on each fruit were duplicated to match the two spectra taken on each fruit. For the stored fruit, SSC0 , DM0 and FF0 predicted from VNIR spectra, measured values of fruit colour (L0 , C0 and H0 ), acoustic firmness (AF0 ) and impact firmness (IPF0 ) at harvest were correlated to FFFinal using stepwise regression. For this analysis, squared values of the fruit attributes measured at harvest were also calculated to take into account possible nonlinear relationships. Data from each cultivar were randomly divided into calibration data set (180 fruit per cultivar) and validation data set (90 fruit per cultivar). Fruit attributes that had significant contribution to reduce
J. Feng et al. / Postharvest Biology and Technology 86 (2013) 17–22
19
Table 1 Calibration and validation statistics for predicting flesh firmness (FF0 ), soluble solids content (SSC0 ) and dry matter content (DM0 ) of apricot fruit using visible-near-infrared (VNIR) spectra. Calibration models were developed separately for each cultivar (130 fruit per cultivar for calibration and 50 fruit per cultivar for validation). SEP = the standard error of prediction. Variablea
Cultivars ‘Clutha Gold’ (LVc = 4)
FF0 (N) ‘Genevieve’(LV = 7) ‘Clutha Gold’(LV = 6) SSC0 (%) ‘Genevieve’(LV = 5) ‘Clutha Gold’(LV = 5) DM0 (%) ‘Genevieve’(LV = 3) a b c
Data set
R2
SEP
Bias
Mean (Std)b
Calibration Validation Calibration Validation
0.79 0.68 0.84 0.84
5.7 5.7 4.7 5.3
0.0 1.7 0.0 1.1
Calibration Validation Calibration Validation
0.73 0.75 0.77 0.77
0.68 0.52 1.02 1.07
0.00 0.09 0.00 −0.32
10.1 (1.3) 10.1 (1.0) 15.2 (2.1) 14.9 (2.2)
Calibration Validation Calibration Validation
0.84 0.74 0.87 0.90
0.55 0.54 0.65 0.64
0.00 0.06 0.00 −0.06
11.3 (1.4) 11.1 (1.1) 16.2 (1.8) 16.2 (1.9)
48.1 (11.8) 50.0 (9.8) 26.5 (11.8) 27.5 (12.7)
“0” in the variable names indicate that the variables were values at harvest (before storage). Mean and standard deviation (Std) of the fruit in each data set. LV stands for the number of latent variables used in the calibration model.
sum squared error of the regression at P < 0.1 significance level were selected. Calibration and validation results were evaluated on the basis of R2 (coefficient of determination), bias (the average difference between actual and predicted values) and SEP (the standard error of prediction). Further analysis of the correlation between FF0 and FFFinal or SSC0 was carried out using a non-linear regression procedure. All the data analysis was performed using Statistical Analysis System, version 9.1 (SAS Inc., USA).
Table 2 Summary of stepwise regression between apricot fruit attributes measured at harvest and the final flesh firmness (FFFinal ) measured after four weeks of refrigerated storage at 0 ◦ C and four days of simulated shelf life (N = 180 fruit for each cultivar). The candidate input variables included acoustic firmness (AF0 ), resonant frequency of the first elliptical mode (Fo0 ), the mass of the fruit (M0 ), and impact firmness (IPF0 ) measured using an Aweta AFS; fruit lightness (L0 ), chroma (C0 ), and hue angle (H0 ) measured using a Minolta Chroma Metre; soluble solids content (SSC0 ), dry matter content (DM0 ) and flesh firmness (FF0 ) estimated from VNIR spectra. Squared values of the original attributes were also used as candidate variables to take account for possible curvilinear relationships. Fruit attributes that had significant contribution to the regression at P < 0.1 significance level were selected. Cultivar
Step
Variable
Estimate
se
Model R2
Pr > F*
1 2 3 4 5 6
Intercept FF0 2 IPF0 2 H0 2 AF0 AF0 2 FF0
−10.5 0.018 0.005 0.002 1.2 −0.03 −1.2
16.2 0.007 0.002 0.001 0.4 0.01 0.7
0.51 0.57 0.60 0.61 0.61 0.62
<0.0001 <0.0001 0.0006 0.06 0.08 0.08
1 2 3 4 5 6
Intercept FF0 2 FF0 M0 AF0 DM0 DM0 2
70.7 0.008 −0.3 −0.03 0.10 −7.2 0.2
18.8 0.002 0.12 0.02 0.04 2.4 0.07
0.80 0.87 0.88 0.88 0.88 0.89
<0.0001 <0.0001 0.008 0.0395 0.0451 0.0059
3. Results 3.1. Prediction of FF0 , SSC0 and DM0 at harvest using VNIR ‘Clutha Gold’
Actual values measured destructively
Calibration models developed for each cultivar based on 130 fruit per cultivar performed well for the validation fruit (Table 1 and Fig. 1). The prediction R2 values for validation data sets were 0.68–0.84, 0.75–0.77 and 0.74–0.90 for FF0 , SSC0 and DM0 , respectively. The corresponding SEP were 5.3–5.7 N for FF0 , 0.52–1.07% for SSC0 and 0.54–0.64% for DM0 . Prediction bias for FF0 ranged from 1.1 to 1.7 N, and the bias for SSC0 and DM0 were −0.32% to 0.09% and −0.06% to 0.06%, respectively (Table 1). 70
A 'Clutha Gold' 'Genevieve'
B
C
20
60
18
50 16
40 30
14
20
12
10 FF0 (N) 0
SSC0 (%)
DM0 (%)
‘Genevieve’
* Probability of a variable being selected into the regression model by random error.
3.2. Predicting FFFinal based on fruit attributes at harvest Stepwise regression between fruit attributes measured at harvest (AF0 , Fo0 , M0 and IPF0 measured using an Aweta AFS; L0 , C0 , and H0 measured using a Minolta Chroma Metre; SSC0 , DM0 and FF0 estimated from VNIR spectra) and FFFinal yielded predictive models involving six selected attributes with calibration R2 values of 0.62 and 0.89 for ‘Clutha Gold’ and ‘Genevieve’, respectively (Table 2).
10
0 15 30 45 60 75 8 10 12 14 16 18 20 22 10 12 14 16 18 20 22 Predicted values from VNIR
Fig. 1. Match between actual flesh firmness (FF0 , A), soluble solids content (SSC0 , B) and dry matter content (DM0 , C) of fruit from two apricot cultivars and the predicted values based on the Visible-near-infrared (VNIR) calibration models established for each cultivar. Each point is the average of two readings measured on each of the 50 validation fruit. Solid lines represent perfect match between VNIR predicted and actual values.
Table 3 Validation statistics of the models to predict final flesh firmness (FFFinal ) using selected apricot fruit attributes measured at harvest (see Table 2 for details). There were 90 fruit of each cultivar used for validation. Cultivar
R2
SEP (N)
Bias (N)
Mean ± Std (N)
‘Clutha Gold’ ‘Genevieve’
0.55 0.78
7.8 2.2
−0.3 0.1
17.4 ± 11.8 5.9 ± 4.7
SEP, the standard error of prediction.
J. Feng et al. / Postharvest Biology and Technology 86 (2013) 17–22
40
40
30
30
20
20
10
10
0
0
0
15 30 45 60 75
100
50
B. 'Genevieve'
15 30 45 60 75
FF0 (N) Fig. 2. Relationship between apricot flesh firmness (FF0 ) estimated from visiblenear-infrared spectra at harvest and flesh firmness measured at the end of four weeks of refrigerated storage at 0 ◦ C and four days of simulated shelf life at 20 ◦ C (FFFinal ). Each point represents a fruit. Solid lines represent predicted values of the fitted exponential model (details in Table 4). Dashed lines represent 95% confidence intervals of the predicted values. The dash-dot line represents the minimum firmness (10 N) required for apricots on the retail shelf. Arrows point to the minimum FF0 values to ensure >97.5% of fruit are firm enough for sale at the end of four weeks of refrigerated storage and four days of shelf life.
When the calibration models derived from the stepwise regression (Table 2) were used to predicted FFFinal of the validation fruit, FFFinal of ‘Clutha Gold’ was predicted with an R2 of 0.55 and a SEP of 7.8 N. FFFinal of ‘Genevieve’ was predicted with an R2 of 0.78 and a SEP of 2.2 N (Table 3). The prediction models for both ‘Clutha Gold’ and ‘Genevieve’ indicated that FF0 was the predominant factor determining FFFinal , as FF0 2 was the first variable entered the regression model (Table 2). Plots of FF0 and FFFinal for each cultivar indicated that the relationship approximated an exponential model (Fig. 2). According to the exponential models fitted for each cultivar (Table 4), ‘Clutha Gold’ apricots needed to be harvested at a FF0 >56 N to ensure >97.5% fruit were firm enough for sale at the end of four weeks of refrigerated storage and four days of shelf life (Fig. 2A). ‘Genevieve’ apricots could be harvested at lower firmness (>47 N) to meet the requirement for firmness at the end of four weeks of refrigerated storage and four days of shelf life (Fig. 2B). To demonstrate the usefulness for segregating and marketing apricots according to firmness at harvest, fruit of each cultivar were graded into five classes of equal fruit numbers according to FF0 estimated from VNIR at harvest. The incidence of fruit that developed visible disorders (soft patches, chilling injury and rots that render fruit unsalable) at the end of four days of simulated shelf life at 20 ◦ C varied considerably among different firmness classes. All cultivars showed an increased incidence of disorders for softer Table 4 Parameters of an exponential model: FFFinal = A·ek·FF0 + c, where FFFinal is flesh firmness measured at the end of four weeks refrigerated storage at 0 ◦ C and four days of simulated shelf life at 20 ◦ C. FF0 is flesh firmness estimated from visible-nearinfrared spectra at harvest, A is amplitude, k is a rate constant and c an offset. N = 270 fruit for each cultivar. Cultivars
Parameters
Estimate
se
R2
RMSE (N)
Mean ± Std (N)
‘Clutha Gold’
A k ca
1.67 0.049 0
0.29 0.003
0.50
8.8
19.1 ± 12.4
‘Genevieve’a
A k c
0.79 0.061 0.99
0.06 0.001 0.38
0.85
2.2
6.3 ± 5.7
a c value was set to 0 for ‘Clutha Gold’, because estimated values at the initial model fitting were not significantly different from 0.
Incidence of disorders (%)
FFFinal (N)
50 A. 'Clutha Gold'
'Clutha Gold' 'Genevieve'
80 60 40 20 0
10
20
30
40
50
FF0 (N) estimated from VNIR at harvest
60
Fig. 3. Incidence of visible disorders (soft patches, chilling injury and rots that render apricot fruit unsaleable) observed at the end of four weeks of refrigerated storage at 0 ◦ C and four days of simulated shelf life at 20 ◦ C. Fruit of each cultivar were graded into five classes of equal fruit numbers according to flesh firmness (FF0 ) estimated from visible-near-infrared spectra at harvest. Each point represents a firmness class that had 54 fruit.
firmness classes. ‘Clutha Gold’ and ‘Genevieve’ fruit harvested at FF0 less than 40 N had high proportions of disordered fruit (Fig. 3). For both cultivars, SSC0 increased as FF0 declined (Fig. 4). According to the exponential models fitted to the data measured at harvest, ‘Clutha Gold’ apricots harvested with a firmness below 49 N had an average SSC0 above 10% (Fig. 4B), the minimum value required to give acceptable taste (Weaver, person. comm., 2009). ‘Genevieve’ apricots would need to have a FF0 less than 69 N to ensure the required soluble solids content (Fig. 4B). 4. Discussion Successful use of VNIR spectroscopy to predict SSC, DM and FF of apricots has been reported previously (Bureau et al., 2009, 2010; Camps and Christen, 2009). The calibration and validation statistics achieved here (Table 1) are similar to those in the literature. The feasibility to use these calibration models in the future was not fully evaluated. Based on our experience with this VNIR instruments, the best prediction results were produced with calibration models developed for each year using variable fruit. However, it would be possible to develop a generic model for each cultivar using data
Soluble solids content (%)
20
18
A. 'Clutha Gold'
B. 'Genevieve'
16 14 12 10 8 6 0
20 30 40 50 60 70 80 20 30 40 50 60 70 80 Flesh firmness (N)
Fig. 4. Relationship between apricot flesh firmness and soluble solids content at harvest. Fruit of each cultivar were graded into five classes of equal fruit numbers according to FF0 . Each point represents a firmness class that had 36 fruit. Vertical bars represent standard deviation. Dashed lines represent exponential curves fitted to the data. The dash-dot line represents the minimum SSC0 required to achieve acceptable taste.
J. Feng et al. / Postharvest Biology and Technology 86 (2013) 17–22
from multiple years covering a wide range of variable fruit. Once a generic calibration model is established, the model could be used in the future without the need to carry out the entire calibration and validation each year. But it would still be necessary check the performance of the calibration model using small number of new season’s fruit in case there are unexpected factors, such as instrument faults and new features of the fruit that affect the prediction. Usually, a bias correction is sufficient instead of developing a new calibration model (Feng et al., 2011). The novelty of this study is in the correlation of fruit attributes measured non-destructively at harvest with firmness measured after refrigerated storage and simulated shelf life. The results from stepwise regression (Table 2) indicated that FF0 is the predominant factor determining FFFinal with partial R2 of 0.51 for ‘Clutha Gold’ and 0.80 for ‘Genevieve’. The next important factor for ‘Clutha Gold’ was squared value of impact firmness (IPF0 2 ) with a partial R2 of 0.06, followed by squared value of fruit hue (H0 2 ) with a partial R2 of 0.03. The contributions of other attributes were minimal with partial R2 less than 0.01. This implies that there was an advantage to combine attributes measured at harvest using the Aweta AFS and Minolta Chroma Metre in addition to FF0 estimated from VNIR spectra. However, the gains in predicting FFFinal from these additional measurements may not be significant enough to justify the use of multiple sensors. For ‘Genevieve’, the partial R2 for all the other attributes apart from FF0 were less than 0.01. Therefore, use of Aweta AFS and Minolta Chroma Metre in addition to VNIR is not recommended for this cultivar. According to the exponential models fitted to the FF0 and FFFinal data (Fig. 2 and Table 4), ‘Clutha Gold’ apricots need to be harvested at a firmness above 56 N to ensure 97.5% saleable fruit at the end of four weeks of refrigerated storage at 0 ◦ C and four days of simulated shelf life at 20 ◦ C. ‘Genevieve’ apricots can be harvested at a lower firmness (>47 N) for sale after the same storage and shelf experience. The exponential relationship between FF0 and FFFinal can be attributed in large to the impact of biological age or biological shift at harvest on propagation of biological variation during postharvest (Hertog et al., 2004; Tijskens et al., 2007; Van de Poel et al., 2012). Based on the sigmoidal model widely used to describe the basic pattern of fruit softening, the relationship between firmness at a fixed storage time (FFFinal in this case) and the firmness variation at harvest (FF0 , an indication of biological age) could be expressed as Eq. (2) (Tijskens et al., 2007). FFFinal =
FFmax −FFmin
1+(FFmax −PFF0 /PFF0 −FFmin )·ek(FFmax −FFmin )t
+ FFmin
(2)
where t represents storage time (days after harvest), FFmax and FFmin represent the highest and the lowest finite limiting firmness values; k represents the rate of softening during storage. When FFmax , FFmin , k and t are treated as constants for each apricot cultivar, the relationship between FFFinal and FF0 follows an exponential decay pattern that represents the lower half of the sigmoidal model (Tijskens et al., 2007). The correlation between firmness at harvest and incidence of disordered fruit at the end of storage and shelf life (Fig. 3) also supports the necessity to segregate, pack and market fruit based on firmness. For this purpose, VNIR measurement at harvest provides a useful tool for fruit segregation. Sensorial acceptability scores of apricots were found to be determined in great extent by soluble solids content (Infante et al., 2006; Harker et al., 2005). It was also found that soluble solids content in apricots remains almost constant once harvested (Aubert and Chanforan, 2007; Infante et al., 2006). ‘Palsteyn’ apricot harvested at an intermediate stage of maturity (light yellow skin colour) with an average soluble solids content of 9.7% reached an average acceptability (Infante et al., 2008). Similarly, 10% soluble solids content is considered the minimum for harvesting apricots in New Zealand
21
(Weaver, person. comm., 2009) as apricots with 9–10% SSC had significantly lower licking scores (Harker et al., 2005). According to the exponential models fitted to our FF0 and SSC0 data (Fig. 4), ‘Genevieve’ apricots harvested with a firmness value below 69 N had an average SSC0 above 10%. This means that ‘Genevieve’ apricots harvested with firmness between 47 and 69 N could satisfy the requirements of both long-term storage and eating quality. However, ‘Clutha Gold’ apricots harvested at a firmness >56 N would be acceptable for long-term storage in terms of maintaining firmness, but the fruit may not be acceptable to consumers because of low soluble solids content. Further work is required to determine the optimal harvest maturity and realistic storage duration for ‘Clutha Gold’.
Acknowledgements This research was funded by the Ministry of Science and Innovation, New Zealand (contract number C06X0806). We thank Prof. Susan Lurie of Agricultural Research Organisation, Israel for comments on initial version of this manuscript.
References Aubert, C., Chanforan, C., 2007. Postharvest changes in physicochemical properties and volatile constituents of apricot (Prunus armeniaca L.): characterization of 28 cultivars. Journal Agricultural and Food Chemistry 55, 3074–3082. Berardinelli, A., Cevoli, C., Silaghi, F.A., Fabbri, A., Ragni, L., Giunchi, A., Bassi, D., 2010. FT-NIR spectroscopy for the quality characterization of apricots (Prunus armeniaca L.). Journal of Food Science 75, E462–E468. Biondi, G., Pratella, G.C., Bassi, R., 1991. Maturity indexes as a function of quality in apricot harvesting. Acta Horticulturae 293, 667–674. Brown, G.S., Walker, T.D., 1990. Indicators of maturity in apricots using biplot multivariate analysis. Journal of the Science of Food and Agriculture 53, 321–331. Bruhn, C.M., Feldman, N., Garlitz, C., Harwood, J., Ivans, E., Marshall, M., Riley, A., Thurber, D., Williamson, E., 1991. Consumer perceptions of quality – apricots, cantaloupes, peaches, pears, strawberries, and tomatoes. Journal of Food Quality 14, 187–195. Bureau, S., Reich, M., Gouble, B., Renard, C.M.G.C., Bertrand, D., Ruiz, D., Audergon, J.M., Lecomte, A., 2010. Determination of apricot fruit quality using Fourier transform infrared spectroscopy. Acta Horticulturae 862, 565–572. Bureau, S., Ruiz, D., Reich, M., Gouble, B., Bertrand, D., Audergon, J.M., Renard, C.M.G.C., 2009. Rapid and non-destructive analysis of apricot fruit quality using FT-near-infrared spectroscopy. Food Chemistry 113, 1323–1328. Camps, C., Christen, D., 2009. On-tree follow-up of apricot fruit development using a hand-held VNIR instrument. Journal of Food Agriculture and Environment 7, 394–400. Carlini, P., Massantini, R., Mencarelli, F., 2000. Vis–VNIR measurement of soluble solids in cherry and apricot by PLS regression and wavelength selection. Journal of Agricultural and Food Chemistry 48, 5236–5242. Cooke, J.R., 1972. An interpretation of the resonant behavior of intact fruits and vegetables. Transactions of the ASAE 15, 1075–1080. Fan, X., Argenta, L., Mattheis, J.P., 2000. Inhibition of ethylene action by 1methylcyclopropene prolongs storage life of apricots. Postharvest Biology and Technology 20, 135–142. Feng, J., McGlone, A.V., Currie, M., Clark, C.J., Jordan, B.R., 2011. Assessment of yellow-fleshed kiwifruit (Actinidia chinensis Hort16A) quality in pre- and postharvest conditions using a portable near-infrared spectrometer. Hortscience 46, 57–63. Gouble, B., Grotte, M., Reling, P., Bureau, S., Audergon, J.M., 2010. An easy and reliable test to define apricot maturity. Acta Horticulturae 862, 539–542. Harker, F.R., Norquay, C., Amos, R., Jackman, R., Gunson, A., Williams, M., 2005. The use and misuse of discrimination tests for assessing the sensory properties of fruit and vegetables. Postharvest Biology and Technology 38, 195–201. Hertog, M.L.A.T.M., Lammertyn, J., Desmet, M., Scheerlinck, N., Nicolaï, B.M., 2004. The impact of biological variation on postharvest behaviour of tomato fruit. Postharvest Biology and Technology 34, 271–284. Infante, R., Kraemer, F., Luchsinger, L., Meneses, C., Aros, D., 2006. Sensorial postharvest quality evolution in apricot (Prunus armeniaca L.) cultivars ‘Palsteyn’ and ‘Grandir’. Acta Horticulturae 717, 321–325. Infante, R., Meneses, C., Defilippi, B.G., 2008. Effect of harvest maturity stage on the sensory quality of ‘Palsteyn’ apricot (Prunus armeniaca L.) after cold storage. Journal of Horticultural Science and Biotechnology 83, 828–832. McGlone, V.A., Jordan, R.B., 2000. Kiwifruit and apricot firmness measurement by the non-contact laser air-puff method. Postharvest Biology and Technology 19, 47–54.
22
J. Feng et al. / Postharvest Biology and Technology 86 (2013) 17–22
Muramatsu, N., Sakurai, N., Yamamoto, R., Nevins, D.J., 1996. Nondestructive acoustic measurement of firmness for nectarines, apricots, plums, and tomatoes. HortScience 31, 1199–1202. Savitsky, A., Golay, J.E.M., 1964. Smoothing and differentiation of data by simplified least squares procedures. Analytical Chemistry 36, 1627–1639. Steinier, J., Termonia, Y., Deltour, J., 1972. Comments on smoothing and differentiation of data by simplified least squares procedure. Analytical Chemistry 44, 1906–1913.
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 Biology and Technology 45, 204–213. Van de Poel, B., Bulens, I., Hertog, M., Van Gastel, L., De Proft, M.P., Nicolai, B.M., Geeraerd, A.H., 2012. Model-based classification of tomato fruit development and ripening related to physiological maturity. Postharvest Biology and Technology 67, 59–67. Visagie, T.R., 1988. Maturity index of apricots. Deciduous Fruit Grower 38, 237–238.
Contents lists available at ScienceDirect
Postharvest Biology and Technology journal homepage: www.elsevier.com/locate/postharvbio
Segregation of apricots for storage potential using non-destructive technologies Jinquan Feng a,∗ , Jill Stanley b , Mohammed Othman c , Allan Woolf a , Maureen Kosasih d , Shane Olsson a , Graeme Clare a , Nick Cooper c , Xirui Wang e a
The New Zealand Institute for Plant & Food Research Limited, Mt Albert Research Centre, Private Bag 92169, Auckland 1142, New Zealand The New Zealand Institute for Plant & Food Research Limited, Clyde Research Centre, 990 Earnscleugh Road, RD 1, Alexandra 9391, New Zealand c Taste Technologies Ltd., 64-72 Victoria Street, Onehunga, Auckland 1061, New Zealand d Westland Milk Products, 56 Livingstone Street, Hokitika 7810, New Zealand e Rural Science and Technology Development Centre of Shaanxi Province, Xian, China b
a r t i c l e
i n f o
Article history: Received 16 March 2012 Accepted 13 June 2013 Keywords: Storage life Flesh firmness Visible-near infrared spectroscopy Acoustic firmness Impact firmness Fruit colour
a b s t r a c t This study was set up to identify critical maturity indices affecting storage potential of apricots and demonstrate the potential for using non-destructive measurements to segregate harvested crops for sequential marketing. Fruit of two apricot (Prunus armeniaca) cultivars (‘Clutha Gold’ and ‘Genevieve’) were harvested and stored for four weeks at 0 ◦ C followed by four days of simulated shelf life at 20 ◦ C. Fruit colour, acoustic firmness, impact firmness, flesh firmness (FF0 ), dry matter content and soluble solids content measured non-destructively at harvest were correlated to the flesh firmness measured at the end of refrigerated storage and simulated shelf life (FFFinal ) through stepwise regression. The regression models indicated that FF0 is a predominant factor determining FFFinal . According to the exponential model describing the relationship between FF0 and FFFinal , ‘Genevieve’ and ‘Clutha Gold’ could be stored at 0 ◦ C for four weeks if harvested at firmness above 47 or 56 N, respectively. Segregation of harvested crops according to FF0 estimated from VNIR would enable sequential marketing of fruit according to storage potential to reduce fruit loss. © 2013 Elsevier B.V. All rights reserved.
1. Introduction Apricot (Prunus armeniaca L.) is one of the most popular summer fruit grown in temperate climate zone with a total world production of about 3.8 million tonnes (FAO 2009). New Zealand currently exports 1000–1500 tonnes of fresh apricots annually worth $NZ7-9 million, which accounts for about 2% of world trade in fresh apricots. Apricots are climacteric fruit with limited storage life of 1–4 weeks (Fan et al., 2000). Mixed maturity at harvest affects storage life and fruit quality in the market, which causes concern (Bruhn et al., 1991; Fan et al., 2000; Infante et al., 2008). Fruit background colour, soluble solids content (SSC) and flesh firmness (FF) have been considered important maturity indices (Visagie, 1988; Brown and Walker, 1990; Biondi et al., 1991; Aubert and Chanforan, 2007; Infante et al., 2008; Gouble et al., 2010). However, the relative importance of these maturity indices for storage potential has not been established. Traditionally, apricots were harvested and graded based on visual assessment of fruit colour. For example, ‘Palsteyn’ apricots
∗ Corresponding author. Tel.: +64 9 9258618; fax: +64 9 9257001. E-mail address: [email protected] (J. Feng). 0925-5214/$ – see front matter © 2013 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.postharvbio.2013.06.015
with orange-yellow skin colour were considered ripe and providing the highest acceptability, and those harvested at an intermediate stage of maturity (light yellow skin colour) provided an average acceptability, while unripe fruit (greenish skin colour) were not acceptable (Infante et al., 2008). However, visual assessment of fruit colour is subjective and can be affected by environmental factors such as sunlight and temperature. Fruit colour is not a reliable representation of flesh firmness or soluble solid content and therefore is only an approximate indication of maturity at harvest. For long-term storage or transport, fruit should be harvested at the pre-climacteric stage, before they attain their full flavour and colour, and are more tolerant to handling and prolonged cold storage (Aubert and Chanforan, 2007). Therefore, the optimal maturity at harvest for each cultivar should be a balance between eating quality and storage potential. Identifying key factors affecting storage potential, and developing a quantitative model to describe the relationship between maturity indices at harvest and fruit quality at the end of targeted storage duration, are important steps to optimising harvest maturity. Recent developments in non-destructive technologies to measure maturity indices offer the possibility of non-invasive on-line packinghouse screening of colour, SSC and firmness at harvest (Muramatsu et al., 1996; Carlini et al., 2000; McGlone and Jordan,
18
J. Feng et al. / Postharvest Biology and Technology 86 (2013) 17–22
2000; Camps and Christen, 2009; Bureau et al., 2010; Berardinelli et al., 2010). These same technologies provide the opportunity to measure both maturity indices at harvest and fruit quality after storage on the same fruit, and to investigate the relationship between these on an individual fruit basis. It was hypothesised that the relationship between maturity indices at harvest and fruit quality at the end of storage could be revealed more clearly on an individual fruit basis than on a fruit batch basis because of the large differences among individual fruit and the elimination of sample error. In this study, a total of 450 fruit from each cultivar were measured using three types of non-destructive measurement technologies at harvest. FF, SSC and dry matter content (DM) were measured destructively on a subsample of 180 fruit from each cultivar immediately after non-destructive measurements so that to develop calibration models to estimate these attributes for the remaining fruit (270 fruit per cultivar) that were stored for four weeks at 0 ◦ C followed by four days of simulated shelf life at 20 ◦ C. Fruit quality measured at the end of refrigerated storage and simulated shelf were correlated with maturity indices measured non-destructively at harvest on an individual fruit basis. 2. Materials and methods 2.1. Fruit harvest Four hundred and fifty apricot fruit for each of the two cultivars (‘Clutha Gold’ and ‘Genevieve’) were harvested from Central Otago, a major apricot production area in New Zealand. Fruit were graded into three maturity classes based on visual assessment of fruit colour (background colour and percentage of blush) and packed separately into single layer trays with plastic Plix® trays. This separation was aimed to reduce maturity variation within each package to minimise the effect of released ethylene from more mature fruit contributing to the softening of less mature fruit. It also ensured that a range of fruit maturities was represented in the sample. Packed fruit were sent to the Plant & Food Research in Auckland on the same day by airfreight. Fruit were kept at ambient temperature overnight before measurements were made.
Fruit colour was measured on opposite flat sides of each fruit along the equatorial zone using a Minolta Chroma Metre (CR300, Konica Minolta Sensing Americas, Inc., USA) equipped with 6-eliment silicon photocells (detector) operated with D65 light source and calibrated with a white calibration plate. Measurement on each fruit was taken on the blushed side, then the shaded side. Measured values were expressed in L (lightness), C (chroma) and H (◦ hue) colour space. The two readings from each fruit were averaged to give one reading for each fruit (i.e. L0 , C0 and H0 , where “0” indicates that the parameter was measured when storage time = 0 days. This applies to all the fruit attributes measured at harvest). Acoustic firmness and impact firmness were measured using an AWETA AFS (AWETATM Impact & Acoustic Firmness System, Nootdorp, Holland). The AFS is a small bench top unit that gently taps the fruit and ‘listens’ to the acoustic response. The vibration pattern (resonance attenuated vibration) of the fruit is analysed and translated into an acoustic firmness (AF0 , Eq. (1)) that is characteristic of the overall stiffness of the fruit (Cooke, 1972). The AFS also measures fruit weight and impact firmness (IPF0 , a measure of the local fruit surface elasticity). Two AFS measurements were made on two flat sides of each fruit along the equatorial plane. The two readings from each fruit were averaged to give a single value for each fruit. 2/3
2.3. Destructive measurements at harvest Destructive measurements of soluble solids content (SSC0 ), dry matter content (DM0 ) and flesh firmness (FF0 ) were made on 180 fruit per cultivar to provide data to develop calibration models to predict FF0 , SSC0 and DM0 for the other 270 fruit per cultivar that went to refrigerated storage and simulated shelf life test. FF0 was measured on two sides of each fruit along the equatorial zone using a 7.9-mm diameter probe attached to a GUSS FTA (GUSS Manufacturing Ltd., South Africa). Penetration speed, trigger force and penetration distance were set to 10 mm s−1 , 50 g and 8 mm, respectively. A patch of skin about 1 mm in thickness was removed from each side of the fruit before firmness measurement. SSC0 was measured using a Digital Hand-Held “Pocket” Refractometer (Model PAL, Atago, Tokyo, Japan) with combined juice expelled from FF0 measurements on each fruit. Dry matter content (DM0 ) was measured using a cross section slice (including skin and flesh, but excluding the stone) cut along the equator of each fruit. The slice was weighed immediately after cutting and then after drying to a constant weight at 65 ◦ C over a period of 24 h. 2.4. Storage and final firmness assessment
2.2. Non-destructive measurements at harvest
AF0 = Fo20 M0
where Fo0 is the resonant frequency of the first elliptical mode (Hz) and M0 is the mass of the fruit (kg). Visible-near-infrared (VNIR) spectra measurements were carried out using a commercial VNIR grading system (Taste Technologies, New Zealand). A continuous beam of light generated from a halogen lamp was focused on fruit passing through the VNIR unit while spectra in a wavelength range of 300–1100 nm with a resolution of 10 nm were taken in a reflectance mode. Each fruit passed through the VNIR unit twice and two VNIR spectra were taken on each fruit. Each plectrum was normalised to reference spectra taken when there were no fruit loaded and transformed into a second derivative spectrum (Savitsky and Golay, 1964; Steinier et al., 1972) before calibration and prediction. These are the most commonly used data transformations necessary to achieve robust calibration models.
(1)
The remaining 270 fruit per cultivar were stored at 0 ◦ C for four weeks and subjected to four days of simulated shelf life at 20 ◦ C under ambient light before the final assessment of flesh firmness (FFFinal ) and disorders such as rots, soft patches and chilling injury that render fruit unsalable. 2.5. Data analysis Calibration models to predict FF0 , SSC0 and DM0 from VNIR spectra were developed based on data from 130 fruit per cultivar. Data from the other 50 fruit per cultivar destructively measured at harvest were used as cross validation data. Calibration models were established using partial least square (PLS) regression between second derivative spectra and the reference data (SSC0 , DM0 and FF0 ). The reference data measured on each fruit were duplicated to match the two spectra taken on each fruit. For the stored fruit, SSC0 , DM0 and FF0 predicted from VNIR spectra, measured values of fruit colour (L0 , C0 and H0 ), acoustic firmness (AF0 ) and impact firmness (IPF0 ) at harvest were correlated to FFFinal using stepwise regression. For this analysis, squared values of the fruit attributes measured at harvest were also calculated to take into account possible nonlinear relationships. Data from each cultivar were randomly divided into calibration data set (180 fruit per cultivar) and validation data set (90 fruit per cultivar). Fruit attributes that had significant contribution to reduce
J. Feng et al. / Postharvest Biology and Technology 86 (2013) 17–22
19
Table 1 Calibration and validation statistics for predicting flesh firmness (FF0 ), soluble solids content (SSC0 ) and dry matter content (DM0 ) of apricot fruit using visible-near-infrared (VNIR) spectra. Calibration models were developed separately for each cultivar (130 fruit per cultivar for calibration and 50 fruit per cultivar for validation). SEP = the standard error of prediction. Variablea
Cultivars ‘Clutha Gold’ (LVc = 4)
FF0 (N) ‘Genevieve’(LV = 7) ‘Clutha Gold’(LV = 6) SSC0 (%) ‘Genevieve’(LV = 5) ‘Clutha Gold’(LV = 5) DM0 (%) ‘Genevieve’(LV = 3) a b c
Data set
R2
SEP
Bias
Mean (Std)b
Calibration Validation Calibration Validation
0.79 0.68 0.84 0.84
5.7 5.7 4.7 5.3
0.0 1.7 0.0 1.1
Calibration Validation Calibration Validation
0.73 0.75 0.77 0.77
0.68 0.52 1.02 1.07
0.00 0.09 0.00 −0.32
10.1 (1.3) 10.1 (1.0) 15.2 (2.1) 14.9 (2.2)
Calibration Validation Calibration Validation
0.84 0.74 0.87 0.90
0.55 0.54 0.65 0.64
0.00 0.06 0.00 −0.06
11.3 (1.4) 11.1 (1.1) 16.2 (1.8) 16.2 (1.9)
48.1 (11.8) 50.0 (9.8) 26.5 (11.8) 27.5 (12.7)
“0” in the variable names indicate that the variables were values at harvest (before storage). Mean and standard deviation (Std) of the fruit in each data set. LV stands for the number of latent variables used in the calibration model.
sum squared error of the regression at P < 0.1 significance level were selected. Calibration and validation results were evaluated on the basis of R2 (coefficient of determination), bias (the average difference between actual and predicted values) and SEP (the standard error of prediction). Further analysis of the correlation between FF0 and FFFinal or SSC0 was carried out using a non-linear regression procedure. All the data analysis was performed using Statistical Analysis System, version 9.1 (SAS Inc., USA).
Table 2 Summary of stepwise regression between apricot fruit attributes measured at harvest and the final flesh firmness (FFFinal ) measured after four weeks of refrigerated storage at 0 ◦ C and four days of simulated shelf life (N = 180 fruit for each cultivar). The candidate input variables included acoustic firmness (AF0 ), resonant frequency of the first elliptical mode (Fo0 ), the mass of the fruit (M0 ), and impact firmness (IPF0 ) measured using an Aweta AFS; fruit lightness (L0 ), chroma (C0 ), and hue angle (H0 ) measured using a Minolta Chroma Metre; soluble solids content (SSC0 ), dry matter content (DM0 ) and flesh firmness (FF0 ) estimated from VNIR spectra. Squared values of the original attributes were also used as candidate variables to take account for possible curvilinear relationships. Fruit attributes that had significant contribution to the regression at P < 0.1 significance level were selected. Cultivar
Step
Variable
Estimate
se
Model R2
Pr > F*
1 2 3 4 5 6
Intercept FF0 2 IPF0 2 H0 2 AF0 AF0 2 FF0
−10.5 0.018 0.005 0.002 1.2 −0.03 −1.2
16.2 0.007 0.002 0.001 0.4 0.01 0.7
0.51 0.57 0.60 0.61 0.61 0.62
<0.0001 <0.0001 0.0006 0.06 0.08 0.08
1 2 3 4 5 6
Intercept FF0 2 FF0 M0 AF0 DM0 DM0 2
70.7 0.008 −0.3 −0.03 0.10 −7.2 0.2
18.8 0.002 0.12 0.02 0.04 2.4 0.07
0.80 0.87 0.88 0.88 0.88 0.89
<0.0001 <0.0001 0.008 0.0395 0.0451 0.0059
3. Results 3.1. Prediction of FF0 , SSC0 and DM0 at harvest using VNIR ‘Clutha Gold’
Actual values measured destructively
Calibration models developed for each cultivar based on 130 fruit per cultivar performed well for the validation fruit (Table 1 and Fig. 1). The prediction R2 values for validation data sets were 0.68–0.84, 0.75–0.77 and 0.74–0.90 for FF0 , SSC0 and DM0 , respectively. The corresponding SEP were 5.3–5.7 N for FF0 , 0.52–1.07% for SSC0 and 0.54–0.64% for DM0 . Prediction bias for FF0 ranged from 1.1 to 1.7 N, and the bias for SSC0 and DM0 were −0.32% to 0.09% and −0.06% to 0.06%, respectively (Table 1). 70
A 'Clutha Gold' 'Genevieve'
B
C
20
60
18
50 16
40 30
14
20
12
10 FF0 (N) 0
SSC0 (%)
DM0 (%)
‘Genevieve’
* Probability of a variable being selected into the regression model by random error.
3.2. Predicting FFFinal based on fruit attributes at harvest Stepwise regression between fruit attributes measured at harvest (AF0 , Fo0 , M0 and IPF0 measured using an Aweta AFS; L0 , C0 , and H0 measured using a Minolta Chroma Metre; SSC0 , DM0 and FF0 estimated from VNIR spectra) and FFFinal yielded predictive models involving six selected attributes with calibration R2 values of 0.62 and 0.89 for ‘Clutha Gold’ and ‘Genevieve’, respectively (Table 2).
10
0 15 30 45 60 75 8 10 12 14 16 18 20 22 10 12 14 16 18 20 22 Predicted values from VNIR
Fig. 1. Match between actual flesh firmness (FF0 , A), soluble solids content (SSC0 , B) and dry matter content (DM0 , C) of fruit from two apricot cultivars and the predicted values based on the Visible-near-infrared (VNIR) calibration models established for each cultivar. Each point is the average of two readings measured on each of the 50 validation fruit. Solid lines represent perfect match between VNIR predicted and actual values.
Table 3 Validation statistics of the models to predict final flesh firmness (FFFinal ) using selected apricot fruit attributes measured at harvest (see Table 2 for details). There were 90 fruit of each cultivar used for validation. Cultivar
R2
SEP (N)
Bias (N)
Mean ± Std (N)
‘Clutha Gold’ ‘Genevieve’
0.55 0.78
7.8 2.2
−0.3 0.1
17.4 ± 11.8 5.9 ± 4.7
SEP, the standard error of prediction.
J. Feng et al. / Postharvest Biology and Technology 86 (2013) 17–22
40
40
30
30
20
20
10
10
0
0
0
15 30 45 60 75
100
50
B. 'Genevieve'
15 30 45 60 75
FF0 (N) Fig. 2. Relationship between apricot flesh firmness (FF0 ) estimated from visiblenear-infrared spectra at harvest and flesh firmness measured at the end of four weeks of refrigerated storage at 0 ◦ C and four days of simulated shelf life at 20 ◦ C (FFFinal ). Each point represents a fruit. Solid lines represent predicted values of the fitted exponential model (details in Table 4). Dashed lines represent 95% confidence intervals of the predicted values. The dash-dot line represents the minimum firmness (10 N) required for apricots on the retail shelf. Arrows point to the minimum FF0 values to ensure >97.5% of fruit are firm enough for sale at the end of four weeks of refrigerated storage and four days of shelf life.
When the calibration models derived from the stepwise regression (Table 2) were used to predicted FFFinal of the validation fruit, FFFinal of ‘Clutha Gold’ was predicted with an R2 of 0.55 and a SEP of 7.8 N. FFFinal of ‘Genevieve’ was predicted with an R2 of 0.78 and a SEP of 2.2 N (Table 3). The prediction models for both ‘Clutha Gold’ and ‘Genevieve’ indicated that FF0 was the predominant factor determining FFFinal , as FF0 2 was the first variable entered the regression model (Table 2). Plots of FF0 and FFFinal for each cultivar indicated that the relationship approximated an exponential model (Fig. 2). According to the exponential models fitted for each cultivar (Table 4), ‘Clutha Gold’ apricots needed to be harvested at a FF0 >56 N to ensure >97.5% fruit were firm enough for sale at the end of four weeks of refrigerated storage and four days of shelf life (Fig. 2A). ‘Genevieve’ apricots could be harvested at lower firmness (>47 N) to meet the requirement for firmness at the end of four weeks of refrigerated storage and four days of shelf life (Fig. 2B). To demonstrate the usefulness for segregating and marketing apricots according to firmness at harvest, fruit of each cultivar were graded into five classes of equal fruit numbers according to FF0 estimated from VNIR at harvest. The incidence of fruit that developed visible disorders (soft patches, chilling injury and rots that render fruit unsalable) at the end of four days of simulated shelf life at 20 ◦ C varied considerably among different firmness classes. All cultivars showed an increased incidence of disorders for softer Table 4 Parameters of an exponential model: FFFinal = A·ek·FF0 + c, where FFFinal is flesh firmness measured at the end of four weeks refrigerated storage at 0 ◦ C and four days of simulated shelf life at 20 ◦ C. FF0 is flesh firmness estimated from visible-nearinfrared spectra at harvest, A is amplitude, k is a rate constant and c an offset. N = 270 fruit for each cultivar. Cultivars
Parameters
Estimate
se
R2
RMSE (N)
Mean ± Std (N)
‘Clutha Gold’
A k ca
1.67 0.049 0
0.29 0.003
0.50
8.8
19.1 ± 12.4
‘Genevieve’a
A k c
0.79 0.061 0.99
0.06 0.001 0.38
0.85
2.2
6.3 ± 5.7
a c value was set to 0 for ‘Clutha Gold’, because estimated values at the initial model fitting were not significantly different from 0.
Incidence of disorders (%)
FFFinal (N)
50 A. 'Clutha Gold'
'Clutha Gold' 'Genevieve'
80 60 40 20 0
10
20
30
40
50
FF0 (N) estimated from VNIR at harvest
60
Fig. 3. Incidence of visible disorders (soft patches, chilling injury and rots that render apricot fruit unsaleable) observed at the end of four weeks of refrigerated storage at 0 ◦ C and four days of simulated shelf life at 20 ◦ C. Fruit of each cultivar were graded into five classes of equal fruit numbers according to flesh firmness (FF0 ) estimated from visible-near-infrared spectra at harvest. Each point represents a firmness class that had 54 fruit.
firmness classes. ‘Clutha Gold’ and ‘Genevieve’ fruit harvested at FF0 less than 40 N had high proportions of disordered fruit (Fig. 3). For both cultivars, SSC0 increased as FF0 declined (Fig. 4). According to the exponential models fitted to the data measured at harvest, ‘Clutha Gold’ apricots harvested with a firmness below 49 N had an average SSC0 above 10% (Fig. 4B), the minimum value required to give acceptable taste (Weaver, person. comm., 2009). ‘Genevieve’ apricots would need to have a FF0 less than 69 N to ensure the required soluble solids content (Fig. 4B). 4. Discussion Successful use of VNIR spectroscopy to predict SSC, DM and FF of apricots has been reported previously (Bureau et al., 2009, 2010; Camps and Christen, 2009). The calibration and validation statistics achieved here (Table 1) are similar to those in the literature. The feasibility to use these calibration models in the future was not fully evaluated. Based on our experience with this VNIR instruments, the best prediction results were produced with calibration models developed for each year using variable fruit. However, it would be possible to develop a generic model for each cultivar using data
Soluble solids content (%)
20
18
A. 'Clutha Gold'
B. 'Genevieve'
16 14 12 10 8 6 0
20 30 40 50 60 70 80 20 30 40 50 60 70 80 Flesh firmness (N)
Fig. 4. Relationship between apricot flesh firmness and soluble solids content at harvest. Fruit of each cultivar were graded into five classes of equal fruit numbers according to FF0 . Each point represents a firmness class that had 36 fruit. Vertical bars represent standard deviation. Dashed lines represent exponential curves fitted to the data. The dash-dot line represents the minimum SSC0 required to achieve acceptable taste.
J. Feng et al. / Postharvest Biology and Technology 86 (2013) 17–22
from multiple years covering a wide range of variable fruit. Once a generic calibration model is established, the model could be used in the future without the need to carry out the entire calibration and validation each year. But it would still be necessary check the performance of the calibration model using small number of new season’s fruit in case there are unexpected factors, such as instrument faults and new features of the fruit that affect the prediction. Usually, a bias correction is sufficient instead of developing a new calibration model (Feng et al., 2011). The novelty of this study is in the correlation of fruit attributes measured non-destructively at harvest with firmness measured after refrigerated storage and simulated shelf life. The results from stepwise regression (Table 2) indicated that FF0 is the predominant factor determining FFFinal with partial R2 of 0.51 for ‘Clutha Gold’ and 0.80 for ‘Genevieve’. The next important factor for ‘Clutha Gold’ was squared value of impact firmness (IPF0 2 ) with a partial R2 of 0.06, followed by squared value of fruit hue (H0 2 ) with a partial R2 of 0.03. The contributions of other attributes were minimal with partial R2 less than 0.01. This implies that there was an advantage to combine attributes measured at harvest using the Aweta AFS and Minolta Chroma Metre in addition to FF0 estimated from VNIR spectra. However, the gains in predicting FFFinal from these additional measurements may not be significant enough to justify the use of multiple sensors. For ‘Genevieve’, the partial R2 for all the other attributes apart from FF0 were less than 0.01. Therefore, use of Aweta AFS and Minolta Chroma Metre in addition to VNIR is not recommended for this cultivar. According to the exponential models fitted to the FF0 and FFFinal data (Fig. 2 and Table 4), ‘Clutha Gold’ apricots need to be harvested at a firmness above 56 N to ensure 97.5% saleable fruit at the end of four weeks of refrigerated storage at 0 ◦ C and four days of simulated shelf life at 20 ◦ C. ‘Genevieve’ apricots can be harvested at a lower firmness (>47 N) for sale after the same storage and shelf experience. The exponential relationship between FF0 and FFFinal can be attributed in large to the impact of biological age or biological shift at harvest on propagation of biological variation during postharvest (Hertog et al., 2004; Tijskens et al., 2007; Van de Poel et al., 2012). Based on the sigmoidal model widely used to describe the basic pattern of fruit softening, the relationship between firmness at a fixed storage time (FFFinal in this case) and the firmness variation at harvest (FF0 , an indication of biological age) could be expressed as Eq. (2) (Tijskens et al., 2007). FFFinal =
FFmax −FFmin
1+(FFmax −PFF0 /PFF0 −FFmin )·ek(FFmax −FFmin )t
+ FFmin
(2)
where t represents storage time (days after harvest), FFmax and FFmin represent the highest and the lowest finite limiting firmness values; k represents the rate of softening during storage. When FFmax , FFmin , k and t are treated as constants for each apricot cultivar, the relationship between FFFinal and FF0 follows an exponential decay pattern that represents the lower half of the sigmoidal model (Tijskens et al., 2007). The correlation between firmness at harvest and incidence of disordered fruit at the end of storage and shelf life (Fig. 3) also supports the necessity to segregate, pack and market fruit based on firmness. For this purpose, VNIR measurement at harvest provides a useful tool for fruit segregation. Sensorial acceptability scores of apricots were found to be determined in great extent by soluble solids content (Infante et al., 2006; Harker et al., 2005). It was also found that soluble solids content in apricots remains almost constant once harvested (Aubert and Chanforan, 2007; Infante et al., 2006). ‘Palsteyn’ apricot harvested at an intermediate stage of maturity (light yellow skin colour) with an average soluble solids content of 9.7% reached an average acceptability (Infante et al., 2008). Similarly, 10% soluble solids content is considered the minimum for harvesting apricots in New Zealand
21
(Weaver, person. comm., 2009) as apricots with 9–10% SSC had significantly lower licking scores (Harker et al., 2005). According to the exponential models fitted to our FF0 and SSC0 data (Fig. 4), ‘Genevieve’ apricots harvested with a firmness value below 69 N had an average SSC0 above 10%. This means that ‘Genevieve’ apricots harvested with firmness between 47 and 69 N could satisfy the requirements of both long-term storage and eating quality. However, ‘Clutha Gold’ apricots harvested at a firmness >56 N would be acceptable for long-term storage in terms of maintaining firmness, but the fruit may not be acceptable to consumers because of low soluble solids content. Further work is required to determine the optimal harvest maturity and realistic storage duration for ‘Clutha Gold’.
Acknowledgements This research was funded by the Ministry of Science and Innovation, New Zealand (contract number C06X0806). We thank Prof. Susan Lurie of Agricultural Research Organisation, Israel for comments on initial version of this manuscript.
References Aubert, C., Chanforan, C., 2007. Postharvest changes in physicochemical properties and volatile constituents of apricot (Prunus armeniaca L.): characterization of 28 cultivars. Journal Agricultural and Food Chemistry 55, 3074–3082. Berardinelli, A., Cevoli, C., Silaghi, F.A., Fabbri, A., Ragni, L., Giunchi, A., Bassi, D., 2010. FT-NIR spectroscopy for the quality characterization of apricots (Prunus armeniaca L.). Journal of Food Science 75, E462–E468. Biondi, G., Pratella, G.C., Bassi, R., 1991. Maturity indexes as a function of quality in apricot harvesting. Acta Horticulturae 293, 667–674. Brown, G.S., Walker, T.D., 1990. Indicators of maturity in apricots using biplot multivariate analysis. Journal of the Science of Food and Agriculture 53, 321–331. Bruhn, C.M., Feldman, N., Garlitz, C., Harwood, J., Ivans, E., Marshall, M., Riley, A., Thurber, D., Williamson, E., 1991. Consumer perceptions of quality – apricots, cantaloupes, peaches, pears, strawberries, and tomatoes. Journal of Food Quality 14, 187–195. Bureau, S., Reich, M., Gouble, B., Renard, C.M.G.C., Bertrand, D., Ruiz, D., Audergon, J.M., Lecomte, A., 2010. Determination of apricot fruit quality using Fourier transform infrared spectroscopy. Acta Horticulturae 862, 565–572. Bureau, S., Ruiz, D., Reich, M., Gouble, B., Bertrand, D., Audergon, J.M., Renard, C.M.G.C., 2009. Rapid and non-destructive analysis of apricot fruit quality using FT-near-infrared spectroscopy. Food Chemistry 113, 1323–1328. Camps, C., Christen, D., 2009. On-tree follow-up of apricot fruit development using a hand-held VNIR instrument. Journal of Food Agriculture and Environment 7, 394–400. Carlini, P., Massantini, R., Mencarelli, F., 2000. Vis–VNIR measurement of soluble solids in cherry and apricot by PLS regression and wavelength selection. Journal of Agricultural and Food Chemistry 48, 5236–5242. Cooke, J.R., 1972. An interpretation of the resonant behavior of intact fruits and vegetables. Transactions of the ASAE 15, 1075–1080. Fan, X., Argenta, L., Mattheis, J.P., 2000. Inhibition of ethylene action by 1methylcyclopropene prolongs storage life of apricots. Postharvest Biology and Technology 20, 135–142. Feng, J., McGlone, A.V., Currie, M., Clark, C.J., Jordan, B.R., 2011. Assessment of yellow-fleshed kiwifruit (Actinidia chinensis Hort16A) quality in pre- and postharvest conditions using a portable near-infrared spectrometer. Hortscience 46, 57–63. Gouble, B., Grotte, M., Reling, P., Bureau, S., Audergon, J.M., 2010. An easy and reliable test to define apricot maturity. Acta Horticulturae 862, 539–542. Harker, F.R., Norquay, C., Amos, R., Jackman, R., Gunson, A., Williams, M., 2005. The use and misuse of discrimination tests for assessing the sensory properties of fruit and vegetables. Postharvest Biology and Technology 38, 195–201. Hertog, M.L.A.T.M., Lammertyn, J., Desmet, M., Scheerlinck, N., Nicolaï, B.M., 2004. The impact of biological variation on postharvest behaviour of tomato fruit. Postharvest Biology and Technology 34, 271–284. Infante, R., Kraemer, F., Luchsinger, L., Meneses, C., Aros, D., 2006. Sensorial postharvest quality evolution in apricot (Prunus armeniaca L.) cultivars ‘Palsteyn’ and ‘Grandir’. Acta Horticulturae 717, 321–325. Infante, R., Meneses, C., Defilippi, B.G., 2008. Effect of harvest maturity stage on the sensory quality of ‘Palsteyn’ apricot (Prunus armeniaca L.) after cold storage. Journal of Horticultural Science and Biotechnology 83, 828–832. McGlone, V.A., Jordan, R.B., 2000. Kiwifruit and apricot firmness measurement by the non-contact laser air-puff method. Postharvest Biology and Technology 19, 47–54.
22
J. Feng et al. / Postharvest Biology and Technology 86 (2013) 17–22
Muramatsu, N., Sakurai, N., Yamamoto, R., Nevins, D.J., 1996. Nondestructive acoustic measurement of firmness for nectarines, apricots, plums, and tomatoes. HortScience 31, 1199–1202. Savitsky, A., Golay, J.E.M., 1964. Smoothing and differentiation of data by simplified least squares procedures. Analytical Chemistry 36, 1627–1639. Steinier, J., Termonia, Y., Deltour, J., 1972. Comments on smoothing and differentiation of data by simplified least squares procedure. Analytical Chemistry 44, 1906–1913.
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 Biology and Technology 45, 204–213. Van de Poel, B., Bulens, I., Hertog, M., Van Gastel, L., De Proft, M.P., Nicolai, B.M., Geeraerd, A.H., 2012. Model-based classification of tomato fruit development and ripening related to physiological maturity. Postharvest Biology and Technology 67, 59–67. Visagie, T.R., 1988. Maturity index of apricots. Deciduous Fruit Grower 38, 237–238.