Sage Journals: Discover world-class research

Abstract

This study was conducted to evaluate the ability of near infrared (NIR) spectroscopy to estimate oil content, and per cent of cake, resin and residue in beauty leaf tree (Calophyllum inophyllum L.) kernel samples. Fruits were collected from various geographical locations of tropical Australia (from Rockhampton to Darwin) and air dried before the kernels were manually separated from the fruits. Kernel samples were oven dried, crushed (5–10 mm) and their NIR spectra collected using a Fourier transform (FT) NIR instrument where the same batch of kernels were used to extract oil using a screw press. Calibration models between the NIR spectra and reference data were developed using partial least squares (PLS) regression. The cross-validation statistics including the coefficient of determination (r²) and standard error in cross validation (SECV) were 0.83 (SECV: 2.39%) for oil content, 0.89 (SECV: 2.81%) for cake, 0.88 (SECV: 1.92%) for resin and 0.79 (SECV: 2.15%) for residue, respectively. This research showed that NIR spectroscopy can be used as an alternative, faster and low-cost technique to predict oil content, per cent of cake, resins and residues in various genotypes of beauty leaf tree. Further studies should be carried out to increase the sample size and chemical variation, as well as to evaluate different methods of oil extraction (e.g., solvent extraction) to improve the reliability of the calibration models.

Keywords

Beauty leaf tree biofuel near infrared resin press cake kernel

Introduction

Botanically, the beauty leaf tree (Calophyllum inophyllum L.) fruit is a ‘drupe’, as it contains an outer layer of soft tissue, middle layer of soft shell plus spongy tissue, and an inner layer defined as kernel where the kernel has the highest content of oil.^1,2 However, in the literature, the kernel is mistakeably referred to as seed or nut.^1,2 The seed includes both the kernel and the soft shell along with the spongy tissue (Figure 1, Panel A). It is therefore appropriate to use the term ‘kernel’ (as opposed to seed or nut) when referring to beauty leaf tree as the seed is rarely used during oil extraction.^1,2 The fatty acid methyl esters and blends originating from the oil of these seeds meet most of the biodiesel requirements in the United States (ASTM D 6751), and European Union (EN 14,214).^1,2 Therefore, its economic importance as alternative and sustainable source of feedstock to produce biofuel in tropical and subtropical regions.^1,2

Figure 1.

Kernel and shell of in the beauty leaf tree (Calophyllum inophyllum L.) seed samples analysed. Panel A. Panel B. layers of oil, resin and residue of the beauty leaf tree (Calophyllum inophyllum L.) seed samples after centrifugation.

During the processing of beauty leaf tree fruits, the kernel is separated from the fruit either manually or mechanically.^1,2 The mechanical separation involves the drying of the kernel to a desirable moisture content (approximately 15%) where a screw press is used to extract the oil. In recent years, other methods have been evaluated and utilized to extract oil such as solvent extraction, aqueous enzymatic oil extraction (AEOE), microwave assisted extraction (MAE), accelerated solvent extraction (ASE), and supercritical fluid extraction.^1–6 However, these methods are time consuming, laborious, and involve the use of harmful chemicals.⁴ Furthermore, these methods are also destructive and cause irreversible damage to the kernel samples.⁷ Hence, these methods are not suitable to either screen different genotypes for breeding purposes or for tree improvement programs which require retention of the kernel for future propagation.^1,2

Rapid, non-destructive techniques such as near infrared (NIR) spectroscopy are proposed and used to estimate oil content, fatty acids and lipids in different type of samples (e.g. rapeseed, safflower, olive oil, corn or maize).^8–13 This method is fast, simple, inexpensive, does not require costly hazardous chemicals, and uses low manpower.¹⁴ Most importantly, NIR spectroscopy is a non-destructive technique and could be utilised to predict oil content in a large number of samples (e.g. intact seeds).^8,9 NIR spectroscopy has been widely used in oilseed research for rapid estimation of oil, fatty acids, and other quality properties (e.g. phenolic compounds, moisture, etc.).^{4,8–13,15,16}

NIR spectroscopy works on the principle of irradiating the samples with NIR light (780 to 2500 nm) and recording the reflected or transmitted light to quantify the organic compounds present in the sample.^14,15,17 For the most part, the reflected, or absorbed NIR spectra originates from three hydrogen containing chemical bonds, viz., N-H (found in protein), C-H (generally found in oils and fats) and O-H (in water).^18,19 To extract the information contained in the spectra, different chemometric methods and techniques are applied.^20–22 Various regression models or algorithms have been used to develop relationships between the reference data and the absorbed spectra (e.g. calibration development).^20–22 Techniques such as partial least squares (PLS) regression and principal component regression (PCR) are commonly used.^20–22 Once a calibration model is developed, the NIR is used to predict the corresponding target parameter from the spectral data of unknown samples.^20–22

Modern instruments have incorporated Fourier-Transformation (FT) to NIR instruments (or FTNIR) to reduce the signal to noise ratio.^14,19,23 The use of FTNIR spectroscopy has been reported for the prediction of oil content, fatty acids, moisture, and protein in different plant species such as Brassica napus,^24–26 Camelina sativa,²⁷ Carthamus tinctorius,²³ Olea europaea,¹⁰ Glycine max (soyabeans),²⁸ Paeonia sect Moutan,¹³ Sesamum indicum,²⁹ Silybum marianum,⁷ and Theobroma cacao³⁰ and Zea mays.⁴ However, no studies have been conducted or reported on the utilization of FTNIR spectroscopy to analyse beauty leaf tree samples.

The present study aims to assess the potential of using FTNIR spectroscopy combined with chemometrics to estimate oil content, and the per cent of cake, resin and residue of beauty leaf tree (Calophyllum inophyllum L.) kernel samples that were collected from various biogeographic regions of tropical Australia.

Materials and methods

Samples and sample preparation

Fruit and seed (fruits without the outer skin) samples of beauty leaf tree (Calophyllum inophyllum L.) were collected from 50 individual trees or accessions located at 19 different biogeographic regions of tropical Queensland and Northern Territory (Australia) (see Table 1).^31,32 Fallen fruit samples that were approximately 3 to 6 months-old were collected from each tree, packed in a hessian bag, and stored in a garden shed in Rockhampton for up to 1 year before analysis (average temp. 24 ± 5°C and ambient humidity 55 ± 10%). The air-dried fruit samples were cracked open manually using a rubber-head hammer where the kernel samples were separated from the rest of the fruit. Kernel samples of approximately 200 g were extracted from each accession, crushed into smaller particles (5–10 mm) using a Nutribullet™ blender and oven dried for 24 h at 65°C–70°C.^31,32

Table 1.

Number of individual trees or accessions collected from the different biogeographic regions or provenances of tropical Queensland and Northern Territory.

Number of trees or accessions	Biogeographic region or provenance
13	Bowen
9	Darwin (Northern Territory)
4	Cardwell
2 (from each biogeographic region)	Cairns, Innisfail, Roslyn Bay, Port Douglas, Bingil Bay, Townsville, Rex Lookout, and South Mission Beach
1 (from each biogeographic region)	Mackay, Wonga Beach, Palm Cove, Trinity beach, Cooya Beach, Cook Town, Newell beach, and Midge point

Oil extraction

Oil extraction was carried out using a ‘Komet’ screw press (IBG Monforts Oekotec GmbH & Co. KG, Germany). Initially, the oil press was pre-heated for 5–8 min by attaching the metal heating ring onto the end of the shaft. When the temperature of the shaft reached around 150°C, the heating ring was removed, and the screw press was turned on and the kernels were manually fed through the hopper. The screw conveyor with its forward rotary motion pressurised the kernels and this facilitated squeezing of the oil from the perforated section of the barrel. Then, the squeezed oil was collected in an aluminium tray and transferred to a plastic container. The press cake was pushed as a ribbon, and this was collected separately. Some kernels, particularly those which had dried to below 10% moisture content were squeezed out of the barrel, as fine particles and mixed once with the oil. At the end of the process, the oil was collected from the aluminium tray and centrifuged to separate the oil, resin and residue (Figure 1, Panel B). The initial temperature of the expelled oil was around 90°C–95°C while the expelled cake was around 75°C–95°C. After extracting the oil from an individual accession, the screw press barrel was dismantled, and thoroughly cleaned with 70% ethanol to prevent sample cross-contamination. The extracted oil was centrifuged for 30 min at 1000 rpm to separate oil from the residues that were deposited at the bottom. After centrifuging, a layer of resin was observed between the oil and residue (see Figure 1, Panel B).

The fractions namely oil content (equation (1)), residue (equation (2)), resin (equation (3)), and cake (equation (4)) were separated manually by pipetting, where the weights of each fraction were recorded, and the content as well as the per cent of each of the constituents were calculated using the equations below.

O i l c o n t e n t (%) = \frac{w e i g h t o f t h e e x t r a c t e d k e r n e l o i l (g)}{w e i g h t o f k e r n e l u s e d i n t h e e x t r a c t i o n (g)} x 100

(1)

R e s i d u e (%) = \frac{w e i g h t o f t h e r e s i d u e (g)}{w e i g h t o f k e r n e l u s e d i n t h e e x t r a c t i o n (g)} x 100

(2)

R e s i n (%) = \frac{w e i g h t o f t h e r e s i n (g)}{w e i g h t o f k e r n e l u s e d i n t h e e x t r a c t i o n (g)} x 100

(3)

C a k e (%) = \frac{w e i g h t o f c a k e o b t a i n e d (g)}{w e i g h t o f k e r n e l s a m p l e u s e d i n t h e e x t r a c t i o n (g)} x 100

(4)

Fourier transformation near infrared spectroscopy

The NIR spectra of the air-dried kernel samples were acquired using an Antaris II Near Infra Red Analyzer (Nicolet, Thermo Electron Corporation, Madison, WI, USA) equipped with a indium gallium arsenide (InGaAs) photodiode detector. Before the NIR spectra were acquired, the samples were kept overnight in an oven at 65°C and then transferred to a desiccator to minimise the effect of moisture. During each scan, the kernel samples (approx. 5 g) were placed in an auto-spinning sample cup and spectra acquired with the integrating sphere. Each sample was repacked and scanned three times. Each recorded spectrum of the sample was an average of 32 scans at 8 cm⁻¹ resolution. The spectral data were acquired in absorbance mode with a wavelength ranging from 4000 to 10,000 cm⁻¹. The RESULT 3 SP8 Build 60 software (Thermo Fisher Scientific Inc., USA) was used to control the instrument and to record the FTNIR spectral data. After scanning, the sample cup was cleaned using isopropyl alcohol and dried using paper wipes (Kimwipes®, Kimberley-Clarks worldwide Inc., USA) to prevent cross-contamination between samples.

Spectral data processing and analysis

The NIR data was exported in Excel format (*.xlsx) into The Unscrambler (version X, CAMO, Norway) for data analysis and pre-processing. The NIR spectra were pre-processed using the Savitzky-Golay second derivative (21 smoothing points and second polynomial order) prior to spectra interpretation and chemometric analysis.³³ Partial least squares (PLS) regression models were developed between the NIR and reference data for the prediction of oil content, per cent of cake, resin and residue in the beauty leaf tree samples. The calibration set was used to develop the PLS regression models using full (leave one out) cross validation.^20–22 The sample set was split into two groups, namely calibration (n:100) and validation (n:50) using the Kennard-Stone (KS) algorithm where replicates were distributed between the calibration and validation sets.³⁴

Quantitative PLS models were evaluated using the coefficient of determination in cross validation (r²_CV), the standard error in cross validation (SECV), bias, slope and residual predictive deviation (RPD). RPD is the ratio of standard deviation (SD) and SECV and was used for testing the reliability of the calibration model.^35,36 The RPD value for a given NIR calibration indicated how well the calibration models predict the reference data of the parameter evaluated.^35,36 According to Williams³⁵ a RPD value higher than 3 is considered sufficient for screening.

Results and discussion

The descriptive statistics including the mean, range, standard deviation (SD), and coefficient of variation (CV) for the determination of oil content, per cent of resin, cake and residue in the beauty leaf tree samples analysed are shown in Table 2. It has been observed a wide variation in oil content (CV: 19.6%), as well as in the per cent of resin (CV: 70.9%), cake (CV: 29.6%) and residue (CV: 48.4%) as indicated by both the range and CV values obtained, respectively. The highest CV value was observed for the determination of per cent of resin (CV: 70.9%) while the lowest was observed for the determination of oil content (CV: 19.6%). Overall, the wide variation in composition exhibited by the samples analysed, has been found to be appropriate to develop NIR calibrations for these parameters.

Table 2.

Descriptive statistics for oil content and percent of cake, resin and residue in the beauty leaf tree (Calophyllum inophyllum L.) kernel samples analysed and used to develop the calibration models (n = 100).

	Mean	SD	Range	CV
Oil%	41.9	8.2	12.9–55.9	19.6
Cake%	34.1	10.1	17.7–67.0	29.6
Resin%	6.2	4.4	0.10–20.4	70.9
Residue%	12.6	6.1	2.7–27.7	48.4

n: number of samples; SD: standard deviation; CV: coefficient of variation.

The average and SD of the second derivative FTNIR spectra of the kernel samples analysed is shown in Figure 2. The average FTNIR spectra showed three main absorbance bands between 8550 and 8200 cm⁻¹ associated with C-H₂ bonds, between 7200 cm⁻¹ and 6900 cm⁻¹ associated with O-H (water content) bonds and C-H (aromatic groups) bonds.³⁷ At 5816 cm⁻¹ and 5669 cm⁻¹ these bands are associated with C-H₂ methylene groups.³⁷ The band at 5199 cm⁻¹ is associated with O-H bands associated with water content, at 4863 cm⁻¹ associated with N-H and CONH₂ (e.g. protein content), at 4686 cm⁻¹ associated with C-H and C=O combination bands of lipids, and at 4366 cm⁻¹and 4231 cm⁻¹ associated with C-H combination bands of aromatic groups.³⁷ The main variation (SD) was observed in the wavenumber region between 6000 and 5590 cm⁻¹ associated with differences in C-H, O-H and C=O bands, related to the content of oil in the kernel samples.³⁷

Figure 2.

Average NIR spectrum and standard deviation of the second derivative of the beauty leaf tree (Calophyllum inophyllum L.) seed samples analysed samples analysed using near infrared spectroscopy.

Table 3 shows the PLS cross-validation statistics for the measurement of oil content, and the per cent of cake, resin and residue the kernel samples analysed. The r²_CV and SECV were 0.83 (SECV: 2.39%) for oil content, 0.89 (SECV: 2.81%) for cake, 0.88 (SECV: 1.92%) for resin and 0.79 (SECV: 2.15%) for residue, respectively. In most cases, the cross-validation models explained more than 79% (r²_CV > 0.79) of the variability in the data set. The RPD_cv values for the measurement of oil content, per cent of cake, resin, and residue were 3.4, 3.5, 2.3 and 2.8 respectively. The RPD_cv values showed that the percent of residue cannot be well predicted using FTNIR spectroscopy (RPD value = 2.3). Overall, the cross-validation statistics showed that the PLS regression models can be used as screening tool for high, medium and low per cent of resin (RPD_cv: 2.8), oil content (RPD_cv: 3.4) and per cent cake (RPD_cv: 3.5). In addition to cross-validation, these models were evaluated using the validation set of samples created by applying the KS algorithm (Table 4). Based on the RPD values both oil content and percent of cake can be predicted using NIR spectroscopy.

Table 3.

Cross validation statistics for the measurement of oil content and percent of cake, resin and residue in the beauty leaf tree (Calophyllum inophyllum L.) samples analysed using near infrared reflectance spectroscopy.

	n	r²_CV	SECV	Bias	Slope	RPD_cv	LV
Oil%	100	0.83	2.39	0.01	0.86	3.4	9
Cake%	100	0.89	2.81	0.02	0.90	3.5	6
Resin%	98	0.88	1.92	−0.03	0.88	2.3	7
Residue%	95	0.79	2.15	−0.02	0.82	2.8	8

n: number of samples; r²_CV: coefficient of determination in cross validation; SECV: standard error in cross validation, RPD_cv; residual predictive deviation in cross-validation; LV: latent variables.

Table 4.

Prediction statistics for the measurement of oil content and percent of cake, resin and residue in the beauty leaf tree (Calophyllum inophyllum L.) samples analysed using near infrared reflectance spectroscopy (n: 50).

	SD	r²	SEP	RPD
Oil%	5.5	0.61	3.50	2.3
Cake%	10.1	0.78	4.55	2.2
Resin%	2.78	0.56	3.30	1.3
Residue%	6.20	0.30	6.15	1.0

n: number of samples; r²: coefficient of regression in prediction or validation; SEP: standard error in prediction; RPD: residual predictive deviation (SD/SEP).

The NIR region comprises of the combination of overtones of the fundamental molecular vibrations absorbing in the IR region.^37,38 It is well known that NIR is more sensitive to anharmonic vibrations, such as C‐H, N‐H and O‐H vibrations.³⁷ This information is used during the development of PLS regression models. Therefore, once the model is developed, the PLS loadings should be interpreted in addition to the calibration and prediction statistics.²⁰ Overall, the analysis and interpretation of the loadings obtained from the PLS calibration models can shed light on the main NIR wavenumbers or wavelengths used by the model to measure the chemical parameters in the set of samples analysed.²⁰ Beebe and collaborators³⁹ also indicated that the loadings might also help to detect unusual variables as well as assists in establishing intrinsic dimensionality of the data set.

Different loadings were observed for each of the PLS regression models developed and shown in Figure 3. The highest loadings were observed in the same NIR regions as those observed in the second derivative reported in Figure 2. However, the magnitude of each of the PLS loadings was different depending on the parameter measured (calibration development). Noticeable, the highest loadings for the measurement of per cent cake were inverse to those observed for the prediction of oil content and per cent of resin. The highest loadings for the measurement of percent of residue were also different and shifted from those observed for the prediction of oil content and per cent of resin.

Figure 3.

Partial least square loadings for the measurement of oil, residue, resin and cake percentage in the beauty leaf tree (Calophyllum inophyllum L.) seed samples analysed using near infrared spectroscopy.

The poor cross-validation statistics for the prediction of percent of resins and residue can be explained for various reasons. The percent of residue varies between samples (accessions) depending on the conditions used of extraction in the screw press (e.g. whether the barrel is cold or hot, the pressure imposed via the quantities of the kernel packed into the hopper, among other). The release of residue particles also varies with the water content of the kernel samples. It is noted that seeds that contain less than 8% moisture are difficult to extract oil from (Ashwath N; personal communication) as they crumble in the screw and extrude through the tiny holes of the barrel. Since the kernel samples used in this study were stored for more than a year, and they were collected from the floor of the forest, it is possible that some of the samples had dried out to below 8% moisture content. Such kernels are likely to contain higher proportion of residue than those having between 10 and 15% moisture. Other issues such as pipetting of the oil and other fractions (e.g. moisture content) will also affect the calibration statistics.

The oil content, as well as the per cent of cake of the kernel samples are interdependent parameters and hence their prediction proved to be more reliable, than for the per cent of residue and resin. Although the calibration models may not be used for a precise analysis of the oil content in the kernel sample analysed, they can be used as rapid screening tool to identify trees that contain high, medium and low oil content. The screening of samples based on the non-destructive analysis of the kernel will be invaluable in plant breeding and selection, as they will equip the breeder with the ability to assess large number of seeds (genotypes or accessions) for seed oil content, with the view to selecting superior genotypes. This technique will also be invaluable in agronomic, ecological and phenological studies wherein the variations between kernels of different maturity stages, seasonal variations (e.g. temperature, rainfall) or agronomic conditions such as fertilizers, irrigation or salinity could be readily elucidated using this non-destructive method.^5,40

Conclusion

This research showed that NIR spectroscopy can be used as an alternative, faster and low-cost technique to characterise oil content, per cent of cake, resins and residues in various genotypes of beauty leaf tree. Non-destructive determination of both oil content and per cent of cake of the kernel is of great significance in plant breeding, as the analysed kernel samples will be utilised in producing superior or adapted genotypes. These results can also assist to analyse and evaluate different phenotypes or accessions as well as agronomic responses of beauty leaf tree, as large number of kernels can be screened no-destructively at low-cost. Further studies should be carried out to increase the sample size and chemical variation, as well as to evaluate different methods of oil extraction (e.g. solvent extraction) to improve the reliability of the calibration models.

Footnotes

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

ORCID iDs

Rahul Sreekumar

Daniel Cozzolino

Kerry Walsh

References

Ashwath

Barton

ACT

. Evaluating biodiesel potential of Australian native and naturalised plant species. Rural industries research and development corporation. Canberra, Australia, 2010. https://rirdc.infoservices.com.au/items/10-216

Bhuiya

MMK

Rasul

Khan

MMK

, et al. Optimisation of oil extraction process from Australian native beauty leaf seed. Calophyllum Inophyllum) Ener Procedia 2015; 75: 56–61.

Baskar

Kalavathy

Aiswarya

, et al. Advances in bio-oil extraction from nonedible oil seeds and algal biomass. In: Azad

(ed) Advances in eco-fuels for a sustainable environment. Woodhead Publishing, 2019, pp. 187–210.

Kotyk

Pagel

Deppermann

, et al. High-throughput determination of oil content in corn kernels using nuclear magnetic resonance imaging. J Am Oil Chem Soc 2005; 82: 855–862.

Fadhlullah

Widiyanto

SNB

Restiawaty

. The potential of Nyamplung (Calophyllum inophyllum L.) seed oil as biodiesel feedstock: effect of seed moisture content and particle size on oil yield. Energy Proc 2015; 68: 177–185.

Asfaw

. Basic essential oil extraction techniques and procedures from aromatic plants. J Chromatogr Separ Tech 2022; 13: 1000489.

Koláčková

Růžičková

Gregor

, et al. Quick method (FT-NIR) for the determination of oil and major fatty acids content in whole achenes of milk thistle (Silybum marianum (L.) Gaertn.). J Sci Food Agric 2015; 95: 2264–2270.

Redha

Torquati

Langston

, et al. Determination of glucosinolates and isothiocyanates in glucosinolate-rich vegetables and oilseeds using infrared spectroscopy: a systematic review. Crit Rev Food Sci Nutr 2023; 10: 1–17. DOI: 10.1080/10408398.2023.2198015

Sato

. New estimation method for fatty acid composition in oil using near infrared spectroscopy. Biosci, Biotechnol, Biochem 2002; 66: 2543–2548.

10.

Kavdir

Buyukcan

, et al. Prediction of olive quality using FT-NIR spectroscopy in reflectance and transmittance modes. Biosyst Eng 2009; 103: 304–312.

11.

Kumar

Andy

. Fourier transform-near infrared reflectance spectroscopy calibration development for screening of oil content of intact safflower seeds. Int Food Res J 2013; 20: 759–762.

12.

Kumar

Chauhan

Kumar

. Screening for erucic acid and glucosinolate content in rapeseed mustard seeds using near infrared reflectance spectroscopy. J Food Sci Technol 2010; 47: 690–692.

13.

Huliang

Fangyun

Liping

. Determination of the fatty acid composition in tree peony seeds using near-infrared spectroscopy. J Am Oil Chem Soc 2016; 93: 943–952.

14.

Beć

Grabska

Huck

. Principles and applications of miniaturized near-infrared (NIR) spectrometers. Chem Eur J 2021; 27: 1514.

15.

Zhang

Tan

Chen

, et al. Rapid prediction of Camelina seed oil content using near-infrared spectroscopy. Energy Fuels. 2017; 31: 5629–5634.

16.

Fassio

Restaino

Cozzolino

. Determination of oil content in whole corn (Zea mays L.) seeds by means of near infrared reflectance spectroscopy. Comput Electron Agric 2015; 110: 171–175.

17.

Budić-Leto

Kljusurić

Zdunić

, et al. Usefulness of near infrared spectroscopy and chemometrics in screening of the quality of dessert wine Prosek. Croat J Food Sci Technol 2011; 3: 9–15.

18.

Cozzolino

Kwiatkowski

Dambergs

, et al. Analysis of elements in wine using near infrared spectroscopy and partial least squares regression. Talanta 2008; 74: 711–716.

19.

Zhang

, et al. Review of NIR spectroscopy methods for non-destructive quality analysis of oilseeds and edible oils. Trends Food Sci Technol 2020; 101: 172–181.

20.

Bureau

Cozzolino

Clark

. Contributions of Fourier-transform mid infrared (FT-MIR) spectroscopy to the study of fruit and vegetables: a review. Postharvest Biol Technol 2019; 148: 1–14.

21.

Naes

Isaksson

Fearn

, et al. A user-friendly guide to multivariate calibration and classification. 2nd edition. Chichester: NIR Publications, 2002.

22.

Williams

Dardenne

Flinn

. Tutorial: items to be include in a report on a near infrared spectroscopy project. J Near Infrared Spectrosc 2017; 25: 85–90.

23.

Blanco

Villarroya

. NIR spectroscopy: a rapid-response analytical tool. Trends Anal Chem 2002; 21: 240–245.

24.

Rossato

Prete

Castro

, et al. Predicting rapeseed oil content with near-infrared spectroscopy. Pesqui Agropecu Bras 2013; 48: 1601–1605.

25.

Olivos-Trujillo

Gajardo

Salvo

, et al. Assessing the stability of parameters estimation and prediction accuracy in regression methods for estimating seed oil content in Brassica napus L. using NIR spectroscopy. In: Chilean conference on electrical, electronics engineering, information and communication technologies (CHILECON), Santiago, Chile, 28–30 October 2015.

26.

Sen

Sharma

Kaur

, et al. Near-infrared reflectance spectroscopy calibrations for assessment of oil, phenols, glucosinolates and fatty acid content in the intact seeds of oilseed Brassica species. J Sci Food Agric 2018; 98: 4050–4057.

27.

Zhang

Wang

, et al. Rapid determination of the oil and moisture contents in Camellia gauchowensis Chang and Camellia semiserrata Chi seeds kernels by near-infrared reflectance spectroscopy. Molecules 2018; 23: 2332.

28.

Lee

Cho

B-K

Kim

, et al. Prediction of crude protein and oil content of soybeans using Raman spectroscopy. Sensor Actuator B Chem 2013; 185: 694–700.

29.

Ozdemir

Karaoglu

Dağ

, et al. Assessment of the sesame oil fatty acid and sterol composition with FT-NIR spectroscopy and chemometrics. Turk J Agric For 2018; 42: 444–452.

30.

Teye

Huang

. Novel prediction of total fat content in Cocoa beans by FT-NIR spectroscopy based on effective spectral selection multivariate regression. Food Anal Methods 2014; 8: 945–953.

31.

Sreekumar

Gunaga

Rajesh P

. Pattern of seed germination in different accessions of Calophyllum inophyllum L. Journal contribution. South Gujarat: CQUniversity, 2020. https://hdl.handle.net/10779/cqu.24791349.v1

32.

Sreekumar

Ashwath

Cozzolino

. Biogeographic variability in kernel oil and press cake content of beauty leaf tree (Calophyllum inophyllum L.), as determined by chemical and near-infrared spectroscopy analysis. Sustainability 2023; 15: 5529.

33.

Savitzky

Golay

MJE

. Smoothing and differentiation of data by simplified least squares procedures. Anal Chem 1964; 36: 1627–1639.

34.

Kennard

Stone

. Computer aided design of experiments. Technometrics 1969; 11: 137–148.

35.

Williams

. Implementation of near-Infrared technology. In: Williams

Norris

(eds). Near infrared technology in the agricultural and food industries. St. Paul, Minn: American Association of Cereal Chemist, 2002, pp. 145–169.

36.

Fearn

. Assessing calibrations: SEP, RPD, RER and R². NIR News 2002; 13: 12–14.

37.

Workman

Weyer

. Practical guide to interpretive near-infrared spectroscopy. Boca Raton: CRC Press Taylor and Francis Group, 2008.

38.

Killner

MHM

Rohwedder

JJR

Pasquini

. A PLS regression model using NIR spectroscopy for on‐line monitoring of the biodiesel production reaction. Fuel 2011; 90: 3268–3273.

39.

Beebe

Pell

Seasholtz

. Chemometrics: a practical guide. Chichester: John Wiley and Sons, 1998, p. 348.

40.

Hathurusingha

Ashwath

Subedi

. Variation in oil content and fatty acid profile of Calophyllum inophyllum L. with fruit maturity and its implications on resultant biodiesel quality. Ind Crops Prod 2011; 33: 629–632.

Predicting oil content of Australian beauty leaf tree kernel samples using near infrared spectroscopy combined with chemometrics

Abstract

Keywords

Introduction

Materials and methods

Samples and sample preparation

Oil extraction

Fourier transformation near infrared spectroscopy

Spectral data processing and analysis

Results and discussion

Conclusion

Footnotes

Declaration of conflicting interests

Funding

ORCID iDs

References