NIR spectroscopy to predict the optimal root harvesting date of chicory (Cichorium intybus L.)

Postharvest Biology and Technology 53 (2009) 77–83

Contents lists available at ScienceDirect

Postharvest Biology and Technology journal homepage: www.elsevier.com/locate/postharvbio

The use of Vis/NIR spectroscopy to predict the optimal root harvesting date of chicory (Cichorium intybus L.) Isabelle M. Franc¸ois a,∗ , Els Mariën a , Kristof Brijs b , Pol Coppin c , Maurice De Proft a a

Department of Biosystems, Division of Crop Bio-engineering, Catholic University Leuven, W. De Croylaan 42, B-3001 Heverlee, Belgium Department of Microbial and Molecular Systems, Centre for Food and Microbial Technology, Catholic University Leuven, Kasteelpark Arenberg 20, bus 2463 B-3001 Heverlee, Belgium c Department of Biosystems, Division M3-Biores, Catholic University Leuven, Kasteelpark Arenberg 20, bus 2300, B-3001 Heverlee, Belgium b

a r t i c l e

i n f o

Article history: Received 21 August 2008 Accepted 7 March 2009 Keywords: Chicory Visible/Near Infrared spectroscopy Optimal harvest date Multivariate analysis

a b s t r a c t The use of Visible/Near Infrared (Vis/NIR) spectroscopy has been evaluated to determine the optimal harvesting date of chicory roots. Indirect maturity parameters, such as the foliage chlorophyll content, root dry matter, sugar and nitrogen content, were measured for two hybrids during the field growing season. The root harvest date, which resulted in the highest percentage of top quality chicory heads after forcing, was chosen as the optimal harvest date. A principal component analysis indicated the presence of both a time and hybrid effect. Partial least square models were constructed to predict the indirect maturity parameters and the number of days before optimal root harvest. Correlations for cross validation between 0.75 and 0.81 were obtained for the prediction of foliage chlorophyll content. For the first time, leaf spectra were used to predict chicory root characteristics with success: root dry matter percentage had a cross validation correlation of 0.81 for the ‘Mont Blanc’ and 0.88 for the ‘Vintor’ hybrids. The prediction of root sugar and nitrogen content was poor (cross validation correlation of 0.62 and 0.74). Predicting the number of days before optimal root harvest was done reasonably accurately, with prediction errors of 8.86 d and 10.61 d for ‘Mont Blanc’ and ‘Vintor’, respectively. The visual part of the spectrum is not required in these calibration models. Data of more years and hybrids should be taken into account to make the models more robust and applicable for farmer practice. © 2009 Elsevier B.V. All rights reserved.

1. Introduction Chicory (‘witloof’, Cichorium intybus L. var. foliosum Hegi) is a cold-requiring long-day plant. The first field year is the vegetative growth phase characterized by the production of a fleshy taproot. The second field year is the generative phase (after vernalization in winter) in which the flowering stem is formed and seeds are produced. To produce the eatable leafy vegetable called chicon, roots are harvested at the end of the first growing period when an appropriate stage of root development (maturity) is reached. Next is the floral bud initiation (cold storage/vernalization) and the third step is the accelerated development of the floral stalk (‘pith’) and surrounding basal leaves under etiolating conditions, called forcing. The end product is the white-yellowish conical shaped cluster of leaves, known as the chicon, ‘witloof’ (synonyms: chicory head, endive, Belgian endive, endive head). It has been shown that the physiological status of these roots at harvest is critical for the quality and storage potential of the roots (Claessens, 1993; Neefs,

∗ Corresponding author. Tel.: +32 16 32 24 07; fax: +32 16 32 29 66. E-mail address: [email protected] (I.M. Franc¸ois). 0925-5214/$ – see front matter © 2009 Elsevier B.V. All rights reserved. doi:10.1016/j.postharvbio.2009.03.003

2001; Quenon et al., 2003). Immature roots have the tendency to grow strictly vegetatively during forcing. A low-quality chicon with incomplete closure of the head is the result. Over-ripe roots are characterized by low vitality and the tendency to form side-shoots and a longer floral stem (Neefs, 2001). Roots with inadequate quality will never produce high quality chicons, despite attempts to correct the situation during forcing (Marle and Roux, 1991). This implies that the physiological status of the roots at the time of harvest is of the highest importance (Marle and Roux, 1991; Claessens, 1993). Currently, decisions about harvest time are mainly based on a subjective visual interpretation of crop characteristics such as leaf colour, texture and canopy structure (Coppenolle et al., 2001). Hence the time the roots are harvested is dependent on an assessment and the know-how of the grower (Van Nerum, 1983). Several physico-chemical properties have been tested in the past as a ripeness test for chicory roots. The size of the internal root cavity (Lips, 1993), dry matter percentage (Coppenolle et al., 2001), the ratio leaf/root weight (Sarrazyn et al., 1992) the change in the root sugar contents (Fiala and Jolivet, 1980), and the reducing power of root extracts (Jolivet et al., 1976, cited by Neefs, 2001) have all been investigated, but none of these methods are commonly used to measure suitability for harvesting and/or root maturity. This is

78

I.M. Franc¸ois et al. / Postharvest Biology and Technology 53 (2009) 77–83

mainly because of several disadvantages linked to each of these proposed methods. In general, the techniques are destructive and often time-consuming, limiting the number of plants that can be sampled. A rapid non-destructive method is needed to predict the optimal root harvest time. Peirs et al. (2000) used Visible/Near Infrared (Vis/NIR) reflectance spectroscopy to predict the optimal picking date of apples. The technique has not been used to predict optimal harvest dates of chicory roots. At the end of the first growing season the foliage starts to relocate leaf components to the root, which serves as a storage organ (Neefs et al., 2002). This leads to the hypothesis that leaf and root characteristics are related. The objective of this study was two-fold, (i) to evaluate the potential to measure foliage chlorophyll content and the internal characteristics of the chicory roots by Vis/NIR spectra of the foliage on the field, and (ii) to predict the optimal harvest date of the roots based on these Vis/NIR spectra.

(lyofilisator Hetosicc type SD 2.5 P = 0.05 mbar) and ground up for measurement of sugars and total nitrogen contents.

2. Materials and methods

2.3.3. Nitrogen content Total nitrogen generally represents only 1% of the root dry weight (Vandendriessche et al., 1993). Nevertheless, research on the roots of the perennial weeds chicory, dandelion and leafy spurge, suggests that nitrogen is an important part of the storage metabolism, because nitrogen compounds undergo significant seasonal changes (Cyr and Bewley, 1989; Cyr et al., 1990). According to Marle and Roux (1991), the nitrogen content of the root determines the final quality of the chicory heads. A nitrogen content lower than 1% of the dry matter content is favourable for head quality. More than 1.2% nitrogen increases the sensitivity of the chicory heads to bacteria and to pith deformations (brown pith), which result in a decrease in quality. In chicory, the amount of nitrogen components, such as amino acids and proteins, varies throughout the season. The root accumulates organic nitrogen compounds at the end of the first growing year (Neefs, 2001). A relationship between the foliage and the root N reserves was demonstrated in the defoliation experiments of Neefs et al. (2002). Nitrogen content of the roots was determined by the Dumas combustion method using a C/N analyser (Elementar Vario Max N/CN, Hanau, Germany).

2.1. Plant material The experiments were performed during the growing season of 2006, using the hybrids ‘Mont Blanc’ (Hoquet) and ‘Vintor’ (Nunhems). Chicory plants (Cichorium intybus L. var. foliosum Hegi) were sown in May at the experimental station ‘Nationale Proeftuin voor Witloof’ in Herent (Belgium). On 13 occasions between July and November, 20 individual plants per hybrid were harvested manually. Plants were transported to the laboratory and stored under ambient conditions prior to measurement preventing water loss. All measurements were carried out on the same day. 2.2. Leaf characteristics The physiological status or ‘ripeness’ of the root at the time of harvest is critical for the quality of the chicory head (Claessens, 1993). The natural maturation of the plants takes place at the end of the growing season (October – November). When foliage senescence starts, leaf components are relocated to the roots. A decrease in chlorophyll content and discolouration of the foliage is the result (Neefs et al., 2002). Therefore, monitoring the chlorophyll content could be useful in determining the optimal time to harvest the roots. The chlorophyll content of the foliage was determined according to the protocol of Moran (1982). N,N-dimethylformamide (DMFA) was used for extraction, and absorption of the extract was measured using a Beckman DU-65 spectrophotometer at wavelengths 603, 625, 647 and 664 nm. Chlorophyll content was calculated using the empirical formulae of Moran (1982). 2.3. Root characteristics 2.3.1. Dry matter content Dry matter content has long been linked to root quality. A root is thought to be ready for harvest when the dry matter content exceeds 21%. Nevertheless, Coppenolle et al. (2001) showed that low quality chicory heads could eventuate even when this criterion is met. Hence, the percentage as such does not seem to be a reliable parameter for root ripeness. However, the dry matter content changes throughout the growing season. The mature phase coincides with a marked increase in plant dry matter resulting from high synthesis and accumulation of fructans in the tuberized roots (Ameziane et al., 1997). After drying a root sample of approximately 5 g at 70 ◦ C for 24 h, the percentage dry matter was calculated from the initial fresh and dry weight. The remainder of each root sample was freeze-dried

2.3.2. Carbohydrate content Once roots are harvested, two major types of reserves are available during storage and forcing. Carbohydrates predominate quantitatively as storage reserves forming a primary source of reserve energy for metabolism during cold storage and for the development of the chicory head during forcing (Fiala and Jolivet, 1980; Van den Ende et al., 1996). The changes in concentrations of glucose, fructose and sucrose during growth, storage and forcing were described in detail by Van den Ende et al. (1996). The concentration of free sugars (d-glucose, d-fructose and sucrose) was quantitatively determined using Boehringer Mannheim test kits, based on an enzymatic spectrophotometric determination of the carbohydrates (Neefs, 2001).

2.4. Collection of the Vis/NIR spectra The upper part of the largest leaf of each plant was measured using a Zeiss Corona 45 VISNIR single-beam diode array spectrometer, which measures absorbance in a range between 400 and 1680 nm. The sampling interval was 2 nm for the entire range, whereas the spectral resolution was 3.3 nm and 6 nm for the 400–950 nm interval and the 950–1680 nm interval, respectively. The bundled detecting and source fibres were placed on a black holder under an angle of 45◦ to avoid specular reflection. The light source consisted of a 10 V/18W halogen lamp. A Spectralon (Zeiss Corona) was used to optimize the instrument and collect the white reference measurement. A black cup was used at the same time to collect the dark reference measurement. The only sample preparation consisted of cleaning and drying the leaf with a paper tissue. 2.5. Statistical analysis One-way ANOVA (Tukey’s studentized range test) was performed on the chemical measurements using SASEnterprise Guide 4.1 (SAS Institute Inc., Cary, NC USA). The reflectance spectra were analysed with the statistical program for multivariate calibration The Unscrambler (Version 9.2, CAMO AS, Trondheim, Norway). The software was used to build multivariate calibration models based on principal component analysis (PCA) and partial least square

I.M. Franc¸ois et al. / Postharvest Biology and Technology 53 (2009) 77–83

regression (PLSR). The calibration models were validated in two different ways. Cross validation was used, which means the dataset was randomly divided into a number of segments. For each regression calculus, one segment was left out of the model development and was used to test its predictive ability. The process was repeated until all segments are tested, and the root mean square error of cross validation (RMSECV) was calculated as the root of the squared average deviation between predicted and measured Y-values in the validation segments (Saevels et al., 2003). Outliers were deleted from the analysis based on their scores, leverage (distance to the model centre for each object summarized over all components) and residuals. A maximum of 5% of the total dataset was considered an outlier, as extensive pruning of the dataset should be avoided (Martens and Naes, 1998). Secondly, as a variant on cross validation, the original dataset was randomly split in a calibration (around 80%) and validation (around 20%) set. The latter was then used to check the validity of the constructed model (Cozzolino et al., 2004). The accuracy of the calibration was here defined by the root mean square error of prediction (RMSEP). To evaluate the predictive ability of the PLS models developed, the Ratio Performance Deviation (RPD), defined as the ratio between the standard deviation of the population (St. Dev.) and the RMSECV for the Vis/NIR calibrations, was also calculated. If the error in estimation for a constituent (RMSECV) is large compared to the spread in composition of that sample in the population (as St. Dev.), then this results in a relatively small RPD. Such calibration models are not considered robust (Moron et al., 2007). The rule of thumb states that a model with a RPD below 1.5 will not lead to reliable results. A RPD above 2 indicates a useful model, above 3 indicates an excellent prediction model (Mouazen et al., 2005).

79

Fig. 1. Change in the chlorophyll content, root dry matter percentage and root nitrogen percentage in the growing season of 2006 for two chicory hybrids. The ‘Mont Blanc’ values are depicted by closed symbols, the ‘Vintor’ values by open symbols. Values are averaged over 10 samples. Standard error bars are shown.

not significantly. Root dry matter percentage followed a logarithmic pattern, while the nitrogen percentage only showed a significant increase at the end of the growing season. The change in glucose, fructose and sucrose concentrations in the root is depicted for both hybrids in Fig. 2 (a, ‘Mont Blanc’, b, ‘Vintor’). Both hybrids showed a similar pattern in the sugar concentrations. The glucose concentrations increased until approximately 100 d after sowing (DAS) and decreased from that point on. Fructose concentrations showed a fluctuating pattern, while sucrose concentrations increased from 128 DAS.

3. Results 3.1. Exploratory analysis of the data For all characteristics in this study, the range, mean value and coefficient of variation, defined as the ratio of the standard deviation and the mean value of the component, are summarized in Table 1. The nitrogen and dry matter content of the roots was significantly higher for ‘Vintor’ than for ‘Mont Blanc’, while the chlorophyll content and the concentration of fructose and sucrose was significantly higher for ‘Mont Blanc’ at the 5% significance level (Tukey’s studentized range testing). Fig. 1 shows the change in chlorophyll content, root dry matter and nitrogen content measured during the growing season. For both hybrids, the chlorophyll content decreased slightly, although

Table 1 Range, mean value, standard deviation (St. Dev.) of the foliage chlorophyll content, root dry matter, nitrogen percentage and sugar content for two chicory hybrids. Hybrid Leaf chlorophyll content (␮g g−1 )

Range

Mean

St. Dev.

Mont Blanc

0.94–3.06

2.01

0.49

Vintor

0.37–3.40

1.78

0.48

22.34 21.67

3.15 2.45

0.68 0.98

0.17 0.28

Root dry matter (%)

Mont Blanc Vintor

16.83–30.26 17.10–27.52

Root nitrogen (%)

Mont Blanc Vintor

0.43–1.23 0.54–1.85

Glucose (␮mol gdw−1 )

Mont Blanc Vintor

3.03–174.25 2.68–226.48

46.15 49.91

43.71 52.37

Fructose (␮mol gdw−1 )

Mont Blanc Vintor

3.92–91.94 4.56–103.41

26.26 21.61

16.97 15.28

Sucrose (␮mol gdw−1 )

Mont Blanc Vintor

41.21–504.68 39.31–300.74

217.22 162.57

82.81 52.12

Fig. 2. Change in the glucose, fructose and sucrose concentration for (a) the Mont Blanc (MB) and (b) the Vintor (VT) chicory hybrid in the growing season of 2006. Values are averaged over 10 samples. Standard error bars are shown.

80

I.M. Franc¸ois et al. / Postharvest Biology and Technology 53 (2009) 77–83 Table 3 PLS models for root dry matter percentage for the ‘Mont Blanc’ (MB) and ‘Vintor’ (VT) hybrid. Hybrid

Validation method

400–1600 nm Cross val

700–1600 nm Cross val

400–1600 nm Test set

MB

Ncal Rval RMSECV/RMSEP LV RPD

320 0.88 1.52 12 2.07

320 0.84 1.71 5 1.84

220 0.83 1.82 9 1.73

VT

Ncal Rval RMSECV/RMSEP LV RPD

330 0.81 1.29 14 1.90

330 0.76 1.42 10 1.73

230 0.73 1.59 7 1.54

Table 4 PLS models for root nitrogen percentage for the ‘Mont Blanc’ (MB) and ‘Vintor’ (VT) hybrid. Fig. 3. PCA of the leaf spectra (400–1600 nm) of the ‘Mont Blanc’ and ‘Vintor’ chicory hybrids in 2006. The first principal component (PC) accounts for 99 % in the total variation in the spectra. Both hybrids are clearly separated from each other.

Fig. 3 depicts the score plot of the first two principal components (PC1 and PC2) for both hybrids based on the leaf spectra (400–1600 nm). It is clear that ‘Mont Blanc’ and ‘Vintor’ were separated along the first principal component, explaining 99% of the variation in the spectra. Calibration models need to be built for each hybrid separately.

Hybrid

Validation method

400–1600 nm Cross val

700–1600 nm Cross val

400–1600 nm Test set

MB

Ncal Rval RMSECV/RMSEP LV RPD

120 0.62 0.13 5 1.31

120 0.66 0.13 5 1.31

80 0.63 0.16 4 1.06

VT

Ncal Rval RMSECV/RMSEP LV RPD

130 0.74 0.18 4 1.56

130 0.66 0.21 5 1.33

90 0.56 0.27 3 1.04

3.2. Predictive models for leaf components It was first attempted to establish a calibration model for the chlorophyll content of chicory leaves. In Table 2, the hybrid, the number of samples to create the model (ncal), the validation correlation (Rval), the RMSECV or RMSEP, the number of latent variables (LV) and the RPD are presented for three kinds of models (spectral range used 400–1600 nm, spectral range between 400 and 700 nm and spectral range used 400–1600 nm with a test set validation). Chlorophyll content was predicted with approximately the same accuracy for ‘Mont Blanc’ (Rval = 0.79, RMSECV = 0.29 g gfw−1 ) as for ‘Vintor’ (Rval = 0.73, RMSECV = 0.27 g gfw−1 ). Correlation and error of prediction values were hardly affected when the NIR range of the spectrum (700–1600 nm) was removed, or a test set validation was used. RPD values ranged between 1.44 and 2.09. The highest values were obtained using the test set validation. 3.3. Predictive models for root components Calibration models were established for the root dry matter percentage, using different validation methods and spectral ranges.

The results are summarized in Table 3. It is clear from these results that the calibration models are hybrid-dependent. For ‘Vintor’, the predictive power for all models is rather high (RMSECV/RMSEP of 1.29–1.59%) compared to the models for ‘Mont Blanc’ (RMSECV/RMSEP of 1.52–1.82%). For both hybrids removing the visual part of the spectrum (400–700 nm) resulted in a decrease of the correlation coefficient and in an increased root mean square error both in calibration and in validation. The models established using an external validation set of 100 samples approached the same results as the models based on cross validation. The results of the calibration models for nitrogen content in the roots are summarized in Table 4. It is clear that no satisfactory model could be established. Correlation of validation is no higher than 0.74. RPD values ranged between 1.04 and 1.56. Table 5 summarizes the results of the PLS models for the glucose, sucrose and fructose content of the roots, as predicted by the spectrum of the green leaf. No satisfying models could be built for the individual sugar content of the chicory roots. The RPD values were all lower than 2. 4. Discussion

Table 2 PLS models for foliage chlorophyll content for the ‘Mont Blanc’ (MB) and ‘Vintor’ (VT) hybrid. Hybrid

Validation method

400–1600 nm Cross val

400–700 nm Cross val

400–1600 nm Test set

MB

Ncal Rval RMSECV/RMSEP LV RPD

320 0.79 0.29 5 1.69

320 0.76 0.30 6 1.63

220 0.80 0.28 2 1.75

VT

Ncal Rval RMSECV/RMSEP LV RPD

330 0.73 0.27 7 1.78

330 0.72 0.27 7 1.78

230 0.80 0.23 4 2.09

4.1. Exploratory analysis of the data The trends noticed in the measured plant characteristics (see Figs. 1 and 2) during a growing season are similar to those described by Claessens (1993) and Neefs (2001). The leaf chlorophyll content has been known to decrease significantly 100 DAS. Leaf components are relocated to the tap root, causing a decrease in chlorophyll content and yellowing of the leaves (Neefs, 2001). However, as there was an exceptionally high mean temperature of 18.4 ◦ C in September (KMI, 2006), this ripening process was less pronounced. The decrease in chlorophyll content was not significant for both hybrids. A rule of thumb states that roots should have a dry matter percentage of at least 20% at harvest. This percentage is nor-

I.M. Franc¸ois et al. / Postharvest Biology and Technology 53 (2009) 77–83

81

Table 5 PLS models for glucose, fructose and sucrose content (␮mol gdw−1 ) of the root for the ‘Mont Blanc’ (MB) and ‘Vintor’ (VT) hybrid. Models were built with the entire spectral range (400–1600 nm) and validated using cross validation. Hybrid

Validation method

Glucose (␮mol gdw−1 )

Fructose (␮mol gdw−1 )

Sucrose (␮mol gdw−1 )

MB

Ncal Rval RMSECV LV RPD

123 0.71 30.99 5 1.41

122 0.61 11.80 5 1.44

121 0.56 62.78 6 1.32

VT

Ncal Rval RMSECV LV RPD

139 0.65 38.77 6 1.35

135 0.34 8.74 3 1.75

138 0.55 40.34 8 1.29

mally reached between 80 and 90 DAS (Ministerie van Landbouw, 1993). However in 2006 it was reached only 110 DAS, possibly due to the exceptional weather conditions (exceptionally hot in June and July, but exceptionally wet in August). Of course, the choice of hybrid used can also influence the change in dry matter percentage. Vandendriessche et al. (1993) performed nitrogen measurements on roots of different fertilization treatments during one growing season. The results (N content between 0.62% and 1.02% for the material with 0 and 40 kg N fertilization) are similar to those in this research. A slight increase with field time was also noticed. Sugars provide energy and structural components for the development of the chicory head (Coppenolle et al., 2001). The glucose, fructose and sucrose levels of the two hybrids showed similar patterns as described by Van den Ende et al. (1996) and Neefs (2001). The former reported the glucose level to decrease from August to September and then to remain quite constant thereafter. In contrast, fructose levels were found to increase five-fold from early October. Our results show a tripling in the fructose concentration from the end of September to the first half of November. Ernst et al. (1995) showed increasing fructose levels in chicory at the onset of autumn. The sucrose level rose significantly from the beginning of September in our research in accordance with other studies on chicory (Ernst et al., 1995; Koch et al., 1999). This increase in sucrose coincides with the breakdown of storage carbohydrates in autumn, which is necessary for the plants to convert high molecular storage carbohydrates into sucrose to increase cold resistance (Koch et al., 1999). 4.2. Predictive models for leaf components Pigments are integrally related to the physiological function of leaves. Because chlorophyll tends to decline more rapidly than carotenoids during leaf senescence (Datt, 1998), we attempted to measure and predict the chlorophyll concentration of chicory leaves using Vis/NIR spectroscopy. Using only the visual part of the spectrum had hardly any impact on the calibration models for ‘Mont Blanc’, and no impact for ‘Vintor’. This implies that a cheaper spectrometer can also be used to measure and predict the chlorophyll content of green chicory leaves. The correlations between measured and predicted values, ranging from 0.73 to 0.80 for both hybrids, are similar to those obtained Datt (1998) and Curran et al. (1992) on fresh Eucalyptus sp. and Amaranthus tricolour leaves, respectively. Correlations between measured and predicted values ranged from 0.78 to 0.91. The RPD values obtained in this study showed that only an intermediate accuracy could be achieved by the Vis/NIR method (RPD < 3). Similarly, Moron et al. (2007) used Vis/NIR spectroscopy to assess chlorophyll in whole-wheat plants. Values of RPD of around 1.0 suggested that no accurate models could be established between the spectra and the chlorophyll measurements.

However, results from different plant species cannot be compared simply. Structural differences between species (i.e. leaf thickness, density, number of air water interfaces, cuticle thickness, and pubescence) do have significant effects on the relationship between leaf reflectance and pigment content (Sims and Gamon, 2002). 4.3. Predictive models for root components 4.3.1. Dry matter percentage Soluble solids and dry matter have been measured in a wide variety of agricultural products using NIRS (Peirs et al., 2000). Van Dijk et al. (2002) reported a correlation of 0.93 between the dry matter percentage and the reflectance spectra of peeled and ground potatoes. However, the use of unpeeled potatoes in the research of Walsh et al. (2004) resulted in a correlation of only 0.79 and a RPD value of 1.6. In this research, the correlation between measured and predicted root dry matter percentage ranged between 0.73 and 0.88. The PLS model using the full spectral range for the ‘Mont Blanc’ hybrid reached the highest RPD value (RPD = 2.07). The RPD values obtained showed that the Vis/NIR method can be a rough screening tool for the dry matter percentage—a high, medium and low scale might be developed. However, in order to develop a robust calibration, more samples from different growing seasons might be necessary. 4.3.2. Individual sugar content Good model performance for soluble solids and dry matter is consistently reported for apple and reasonable performance for mandarin, onion and mango (Walsh et al., 2004). NIR from 800 to 1000 nm has been used for predicting the content of fructose and glucose in kiwifruit. The RMSEC values were 1.96% fructose and 1.68% glucose (Slaughter and Crisosto, 1998). However, when predicting sugar content of tubers and roots, results are less in agreement. Scanlon et al. (1999) concluded that NIR in the 770–2498 nm region was unable to act as a screening tool for unacceptable sugar content in potatoes, since no acceptable model could be built. A possible explanation for this difficulty was the lower concentration of sugars in potato, compared to the concentrations in fruit. In contrast, Diller (2002) used the wavelength region between 1100 and 2500 nm to generate successfully partial least square (PLS) models for the prediction of glucose, fructose and sucrose content (RMSEP of 0.7–1.0 mg gfw−1 ) of potato. Calibration models for the sugar content in carrots have been established using reflectance measurements in the range of 800–1700 nm. Prediction errors ranged from 0.51 mg gfw−1 for fructose to 2.86 mg gfw−1 for sucrose (Zude et al., 2007). All spectral measurements in the previously mentioned research were performed on the root or tuber itself. Our research, performed with spectral measurements of the green leaf, resulted in calibration models with RMSECV values of 11.80–62.78 ␮mol g dry weight−1

82

I.M. Franc¸ois et al. / Postharvest Biology and Technology 53 (2009) 77–83

(comparable with 0.47–4.77 mg gfw−1 with an average root dry matter of approximately 22%). De Belie et al. (2003) noted that the SSC of carrots could be estimated using Vis/NIR reflectance measurements with an average RMSEP of 13% of the measured SSC range. In comparison, the different sugar levels could be estimated from the spectra with an RMSECV of between 13 and 18% of the sugar’s measured range. RPD values indicated that the leaf spectra were unable to provide even a rough prediction of root sugar content. 4.3.3. Root nitrogen content Total nitrogen content in chicory roots is important for root storage. A concentration below 1% is preferred (Van Nerum, 1983; Marle and Roux, 1991). This root nitrogen percentage is related to the nitrogen index of the field, fertilization, climate and harvest time and is determinative for the chicon quality (Coppenolle et al., 2001). In the past, NIR techniques have been successfully used to estimate nitrogen concentration in different plants and plant parts (Young et al., 1997). Gislum et al. (2004) measured nitrogen content in dried and ground ryegrass with a correlation of 0.98 and a RMSEP between 0.19 and 0.24%. The nitrogen percentage of whole-wheat plants was also quite well predicted by Moron et al. (2007) (RPD between 1.9 and 2.3). However, after examining the nitrogen content of potatoes, Haase (2006) only obtained RPD values as low as 1.54. The NIR method used could only be applied for rough grouping between tubers high and low in nitrogen content. Only the PLS model for ‘Vintor’ reached a similar RPD value of 1.56. However, this low correlation could be attributed to the limited range and the high variation in root nitrogen content (see Table 1). Therefore, different nitrogen fertilizers should be used in future experiments. It should be noted that all of the spectral measurements in the literature were performed on the fruit or tuber in question for which the models were developed. In the case of chicory, spectral measurements have been performed on a cross section of roots at harvest and linked to characteristics of the chicory head after forcing (Coppenolle et al., 2001). It seemed possible to some extent to predict quality and yield of the chicory heads. However, manually harvesting the roots during the growing season and performing the measurements is still time-consuming and destructive. The potential of spectral measurements at the canopy or leaf level to predict underground biomass of roots has been investigated in a few rare cases (Gat et al., 2000). The accuracy obtained using samples from one growing season and with one fertilization treatment is considered promising for some characteristics. The possibility of measuring root characteristics using a leaf reflectance spectrum should further be investigated as it offers many advantages over destructive methods on one hand, and non-destructive methods on tubers or roots that have to be dug up on the other. Acknowledgements The authors would like to thank the Nationale Proeftuin voor Witloof, Herent for providing the experimental field and producing the chicory heads. We thank Dr. Ir. B. Sm. Devogelaere for comments and discussion. References Ameziane, R., Richard-Molard, C., Deleens, E., Morot-Gaudry, J.F., Limami, A.M., 1997. Nitrate ((NO3)-N-15) limitation affects nitrogen partitioning between metabolic and storage sinks and nitrogen reserve accumulation in chicory (Cichorium intybus L.). Planta 202, 303–312. Claessens, G., 1993. Biometrische en biochemische studie van witloof (Cichorium intybus L.) in funktie van het forceriegedrag. Doctoral Thesis. Catholic University Leuven, Leuven, Belgium. Coppenolle, H., De Baerdemaeker, J., De Proft, M., Hendrickx, M., Leuridan, S., Nicolaï, B., Quenon, V., Schrevens, E., Van de Velde, M., Vanstreels, E., 2001. Integraal

kwaliteitsbeheer bij productie, behandeling en verwerking van witloof. Catholic University Leuven, Leuven, Belgium. Cozzolino, D., Esler, M.B., Dambergs, R.G., Cynkar, W.U., Boehm, D.R., Francis, I.L., Gishen, M., 2004. Prediction of colour and pH in grapes using a diode array spectrophotometer (400–1100 nm). J. Near Infrared Spectrosc. 12, 105–111. Curran, P.J., Dungan, J.L., Macler, B.A., Plummer, S.E., Peterson, D.L., 1992. Reflectance spectroscopy of fresh whole leaves for the estimation of chemical concentration. Rem. Sens. Environ. 39, 153–166. Cyr, D.R., Bewley, J.D., 1989. Carbon and nitrogen reserves of leafy spurge (Euphorbia Esula) roots as related to overwintering strategy. Physiol. Plant. 77, 67–72. Cyr, D.R., Bewley, J.D., Dumbroff, E.B., 1990. Seasonal dynamics of carbohydrate and nitrogenous components in the roots of perennial weeds. Plant Cell Environ. 13, 359–365. Datt, B., 1998. Remote sensing of chlorophyll a, chlorophyll b, chlorophyll a+b, and total carotenoid content in eucalyptus leaves. Rem. Sens. Environ. 66, 111–121. Diller, M., 2002. Untersuchungen zur NIRS-Methodenentwicklung für Kartoffeln aus dem Organischen Landbau unter Berücksichtigung von Jahrgangs-und Sorteneinflüssen. Doctoral Thesis. Rheinische Friedrich-Wilhelms-Universität Bonn, Institut für Lebensmittelwissenschaft und Lebensmittelchemie, Bonn, Germany. Ernst, M., Chatterton, N.J., Harrison, P.A., 1995. Carbohydrate changes in chicory (Cichorium intybus L. var. foliosum) during growth and storage. Sci. Hort. 63, 251–261. Fiala, V., Jolivet, E., 1980. The aptitude of roots of witloof chicory for chicon production studied by their carbohydrate composition. Sci. Hort. 13, 125–134. Gat, N., Erives, H., Fitzgerald, G.J., Kaffka, S.R., Maas, S.J., 2000. Estimating sugar beet yield using aviris-derived indices. In: AVIRIS Earth Science & Applications Workshop. JPL, Pasadena, CA, USA. Gislum, R., Micklander, E., Nielsen, J.P., 2004. Quantification of nitrogen concentration in perennial ryegrass and red fescue using near-infrared reflectance spectroscopy (NIRS) and chemometrics. Field Crops Res. 88, 269–277. Haase, N.U., 2006. Rapid estimation of potato tuber quality by near-infrared spectroscopy. Starch-Starke 58, 268–273. KMI, 2006. Klimatologisch jaaroverzicht van 2006. http://www.meteo.be/meteo/ view/nl/155270-2006.html. Koch, K., Andersson, R., Rydberg, I., Aman, P., 1999. Influence of harvest data on inulin chain length distribution and sugar profile for six chicory (Cichorium intybus L.) cultivars. J. Sci. Food Agric. 79, 1503–1506. Lips, J., 1993. Het rooien van de wortelen. In: de Baerdemaeker, M. (Ed.), Witloofteelt. Ministerie van Landbouw, Belgium, pp. 81–90. Marle, M., Roux, L., 1991. De voedingsoplossing voor het forceren van witloof. XIde Tweejaarlijkse internationale witloofdagen, Arras, France, pp. 89–99. Martens, H., Naes, T., 1998. Multivariate calibration. John Wiley and Sons Ltd., Chichester, UK. Ministerie van Landbouw, 1993. Witloofteelt, Brussels, Belgium. Moran, R., 1982. Formulae for determination of chlorophyllous pigments extracted with N,N-dimethylformamide. Plant Physiol. 69, 1376–1381. Moron, A., Garcia, A., Sawchik, J., Cozzolino, D., 2007. Preliminary study on the use of near-infrared reflectance spectroscopy to assess nitrogen content of undried wheat plants. J. Sci. Food Agric. 87, 147–152. Mouazen, A.M., Karoui, R., De Baerdemaeker, J., Ramon, H., 2005. Classification of soil texture classes by using soil visual near infrared spectroscopy and factorial discriminant analysis techniques. J. Near Infrared Spectrosc. 13, 231–240. Neefs, V., 2001. Veldfase van de witloofwortel (Cichorium intybus L. ) als basis voor bewaring. Doctoral Thesis. Catholic University Leuven, Belgium. Neefs, V., Marechal, I., Hernandez-Martinez, R., De Proft, M.P., 2002. The influence of mechanical defoliation and ethephon treatment on the dynamics of nitrogen compounds in chicory (Cichorium intybus L.). Sci. Hort. 92, 217–227. Peirs, A., Lammertyn, J., Ooms, K., Nicolaï, B.M., 2000. Prediction of the optimal picking date of different apple cultivars by means of VIS/NIR-spectroscopy. Postharvest. Biol. Technol. 21, 189–199. Quenon, V., Jancsok, P., De Baerdemaeker, J., 2003. Optical detection of the internal quality of Belgian endive. In: Verlinden, B., Nicolaï, B., De Baerdemaeker, J. (Eds.), Postharvest. Unlimited Act. Hort., 373–376. Saevels, S., Lammertyn, J., Berna, A.Z., Veraverbeke, E.A., Di Natale, C., Nicolaï, B.M., 2003. Electronic nose as a non-destructive tool to evaluate the optimal harvest date of apples. Postharvest. Biol. Technol. 30, 3–14. Sarrazyn, R., Vanderschelden, C., Coysman, A., Calus, A., Vanongeval, L., Vandendriessche, H., Deckers, S., Geypens, M., 1992. Invloed van de bodemfertiliteit en de stikstofbemesting op de samenstelling van witloofwortelen en het forceerresultaat. Provinciaal Onderzoek-en Voorlichtingscentrum voor Land-en Tuinbouw, Beitem-Rumbeke, en Bodemkundige Dienst van België, Belgium. Scanlon, M.G., Pritchard, M.K., Adam, L.R., 1999. Quality evaluation of processing potatoes by near infrared reflectance. J. Sci. Food Agric. 79, 763–771. Slaughter, D.C., Crisosto, C.H., 1998. Nondestructive internal quality assessment of kiwifruit using near-infrared spectroscopy. Sem. Food Anal. 3, 131–140. Van den Ende, W., Mintiens, A., Speleers, H., Onuoha, A.A., Van Laere, A., 1996. The metabolism of fructans in roots of Cichorium intybus during growth, storage and forcing. New Phytol. 132, 555–563. Van Dijk, C., Fischer, M., Holm, J., Beekhuizen, J.-B., Stolle-Smits, T., Boeriu, C., 2002. Texture of cooked potatoes (Solanum tuberosum). 1. Relationships between dry matter content, sensory-perceived texture, and near-infrared spectroscopy. J. Agric. Food Chem. 50, 5082–5088.

I.M. Franc¸ois et al. / Postharvest Biology and Technology 53 (2009) 77–83 Van Nerum, K., 1983. De teelt van witloofwortelen en de trek van witloof in containers. In: Van Nerum, K. (Ed.), De moderne witloofteelt. Vereniging ‘Het Nieuwe Witloof’ VZW, WTC Tower 1, Brussels, Belgium, pp. 73–76. Vandendriessche, H., Sarrazyn, R., Calus, A., Deckers, S., Geypens, M., Vanderschelden, C., Vanongeval, L., Coysman, A., 1993. Invloed van de bodemfertiliteit en de N-bemesting op de samenstelling van de witloofwortel en het forceerresultaat. In: De Proft, M.P., Van Acker, W.F.J.V. (Eds.), De witloofwortelproduktie, basis van de rendabiliteit. 12de tweejaarlijkse internationale witloofdagen, Leuven, Belgium, pp. 2–12.

83

Walsh, K.B., Golic, M., Greensill, C.V., 2004. Sorting of fruit using near infrared spectroscopy: application to a range of fruit and vegetables for soluble solids and dry matter content. J. Near Infrared Spectrosc. 12, 141–148. Young, M.W., Mackerron, D.K.L., Davies, H.V., 1997. Calibration of near infrared reflectance spectroscopy to estimate nitrogen concentration in potato tissues. Potato Res. 40, 215–220. Zude, M., Birlouez-Aragon, I., Paschold, P.J., Rutledge, D.N., 2007. Non-invasive spectrophotometric sensing of carrot quality from harvest to consumption. Postharvest. Biol. Technol. 45, 30–37.