







































 
 

 

38 
© 2020 by the authors; licensee Asian Online Journal Publishing Group 
 

Agriculture and Food Sciences Research 
Vol. 7, No. 1, 38-45, 2020 

ISSN(E) 2411-6653/ ISSN(P) 2518-0193 
DOI: 10.20448/journal.512.2020.71.38.45 

© 2020 by the authors; licensee Asian Online Journal Publishing Group 

    
 

 
 
 
Developing Calibration Model for Prediction of Malt Barley Genotypes Quality 
Traits using Fourier Transform near Infrared Spectroscopy 

 
Yadesa Abeshu1    
Ashagrie Zewdu2 

 

 
( Corresponding Author) 

 
1Food Science and Nutrition Research Program, Holeta Agricultural Research Center, Ethiopian Institute of 
Agricultural Research; Addis Ababa, Ethiopia. 

 
2Center for Food science and Nutrition, College of Natural and Computational Sciences, Addis Ababa University; 
Addis Ababa, Ethiopia. 

 

 
Abstract 

Ethiopia is the second largest barley producer in Africa. But the quality traits are always 
influenced by the cultivar itself and growing environment. Thus, the study was targeted on 
developing calibration model for predicting malt barley quality traits of genotypes grown at 
different locations using near infra-red spectroscopy for selection purposes in barley breeding 
program. For this purpose, 60 barley samples were collected from Holeta, Debre-Birhan and 
Bekoji. Samples were chemically analyzed in duplicate for 5 barley traits. The calibration model 
was developed based on 120 samples spectral data and 60 chemistry data results using the 
calibration software of the FT-Near Infrared Spectroscopy. The barley Protein calibration model 
having (R²c= 0.97; RPD=5.7 and R2

c=0.94; RPD=4.16) respectively, can be regarded as broadly 
applicable; Extract and Friability (R²c= 0.96; RPD=4.54 and R2

c=0.95; RPD=4.36) respectively 

were accepted as useable with good prediction capability; whereas β-Glucan calibration model 
(R²c= 0.90; RPD=3.18) allowed only for screening purpose in some applications. Barley grain dry 
matter with model parameters result (R2

c=0.86; RPD=2.69)shown usable with caution only for 
rough screening purposes. Hence near infrared spectroscopy is fast and cost-efficient, the breeding 
program can increase the intensity of variety selection using calibration models reflected good 

predicting performances except models for dry matter and β-glucan. 
 

Keywords: Barley, Genotypes, Malt, Trait, Near infrared spectroscopy, Calibration, Validation. 

 
Citation | Yadesa Abeshu; Ashagrie Zewdu (2020). Developing 
Calibration Model for Prediction of Malt Barley Genotypes Quality 
Traits using Fourier Transform near Infrared Spectroscopy. 
Agriculture and Food Sciences Research, 7(1): 38-45. 
History:  
Received: 7 January 2020 
Revised: 10 February 2020 
Accepted: 17 March 2020 
Published: 13 April 2020 
Licensed: This work is licensed under a Creative Commons 

Attribution 3.0 License  
Publisher:  Asian Online Journal Publishing Group 
 

Acknowledgement: Authors would like to express their gratitude to Rebeka 
G/Tsadik (PhD) GIZ contact person and coordinator in Ethiopia and 
BerhaneLakew (PhD) senior barley breeder at Holeta Research Center for 
their kind supports to the successful attainment for the study. Their special 
thanks also go to Wilde Peer (PhD) KWS breeding Company senior breeding 
advisor and Christian Utschig (PhD) for coordinating laboratory facility at 
VLB, Berlin and assisted them in providing NIRS statistical materials. Besides 
they would like to thank Addis Ababa University, Center for Food Science and 
Nutrition as well as Holetta Agricultural Research Center barley breeding 
team members and Food science and Nutrition Research Program for their 
support in sample collection and laboratory analysis. 
Funding: The authors would like to thank the GIZ-SSAP project and KWS 
breeding Company of Germany which supported them in financial and 
technical assistance. 
Competing Interests: The authors declare that they have no conflict of 
interests. 
Transparency: The authors confirm that the manuscript is an honest, 
accurate, and transparent account of the study was reported; that no vital 
features of the study have been omitted; and that any discrepancies from the 
study as planned have been explained. 
Ethical: This study follows all ethical practices during writing.   

 

 

Contents 
1. Introduction ...................................................................................................................................................................................... 39 
2. Materials and Methods ................................................................................................................................................................... 39 
3. Results and Discussion ................................................................................................................................................................... 41 
4. Conclusion ......................................................................................................................................................................................... 44 
5. Recommendation.............................................................................................................................................................................. 45 
References .............................................................................................................................................................................................. 45 
 

http://crossmark.crossref.org/dialog/?doi=10.20448/journal.512.2020.71.38.45&domain=pdf&date_stamp=2017-01-14
http://crossmark.crossref.org/dialog/?doi=10.20448/journal.512.2020.71.38.45&domain=pdf&date_stamp=2017-01-14
http://creativecommons.org/licenses/by/3.0/
http://creativecommons.org/licenses/by/3.0/
https://www.asianonlinejournals.com/index.php/AESR/article/view/1503
http://orcid.org/0000-0002-4648-2370


Agriculture and Food Sciences Research, 2020, 7(1): 38-45 

39 
© 2020 by the authors; licensee Asian Online Journal Publishing Group 

 

 

Contribution of this paper to the literature 
This study was targeted on developing calibration model for predicting malt barley quality traits of 
genotypes grown at different locations using near infra-red spectroscopy for selection purposes in barley 
breeding program. 

 
1. Introduction 

Barley (Hordeum vulgare L.) is the fourth most important cereal crop worldwide after wheat, corn and rice [1]. 
But it is the fifth most important grain crop in Ethiopia and has diverse ecologies being grown from1800 to 3400m 
altitude in different seasons and production systems which makes Ethiopia being the second largest producer in 
Africa, next to Morocco.  Also recognized as one of the country’s most ancient food crop, which is believed to have 
first domesticated about 10,000 years ago from its wild relatives in the Fertile Crescent of the Near East and center 
of diversity in Ethiopia [2]. It is most widely used for the production of malt worldwide [3]. Barley malt also 
mainly used as a source of fermentable sugars for alcoholic fermentation like production of beer [4]. Barley isa 
complex mixture of many organic components that include protein, starch, oil, polysaccharides and sugars. The 
amount of each of the constituents will vary due to both the genetic background and the environmental conditions 
during grain development. The malting process of barley in particular, modifies the grain components during the 
controlled steeping, germination and drying processes. Malt is an essential ingredient in beer production where 
soluble components of the malt are extracted into a liquid broth called wort [5]. The ability to predict grain 
quality for different purposes in early generations would be of great benefit to breeders and industries, allowing for 
selection of suitable lines to deliver product of the highest quality. At later stages in the barley breeding programs, 
micro-malting can be carried which requires large barley sample sizes, destructive and requires experienced 
personnel. For this purpose, near infrared spectroscopy is an ideal technique as it is fast, reliable and non-
destructive which does not require large sample sizes [6]. Near infrared spectroscopy is a type of vibration 
spectroscopy that employs photon energy in the range of 7.96 x 10-20 to 2.65 x 10-19 J. The range is higher than 
necessary to promote molecules to their lowest excited vibrational states and lower than typical values necessary 
for electron excitation in molecules [7]. Hence, food NIR spectra comprised of broad bands arising from 
overlapping absorptions corresponding mainly to overtones and combinations of vibration modes involving C–H, 
O–H, N–H and SH chemical bonds [8]. Thus, sophisticated mathematical techniques, termed chemometrics, are 
heavily employed to allow calibration for reliable extraction of relevant information encoded in the NIR spectral 
data [9]. The aim of this study is to determine constituents of particular interest for the brewing of barley grain 
criterion important traits by using NIRS. This was performed using Near Infra-Red Reflectance (NIR) technology 
which is effective in simultaneously predicting of multiple constituents in agricultural products.  
 

2. Materials and Methods 
2.1. Materials 

The samples used in this study were from barley breeding program trials of Ethiopian Institute of Agricultural 
Research Centers having the mandate of barley growing and breeding potential around central highland areas of 
Ethiopia. Thus, barley samples were collected representing a range of breeding generations fully-fledged at 
different environments throughout barley growing highland areas, specifically from Holeta, Debre Birhan and 
Bekoji locations. About 60 samples from 2018 year of barley growing season were collected from the pre-specified 
growing areas for the study. Particularly, from the plot of the breeding program-controlled trials depending on 
genotypes, location, type (food or malt barley) and quality variability.  
 

2.2. Sample Preparation 
Barley samples which were collected from different locations and different genotypes of breeding trials were 

selected purposively from different plots depending on agronomical data and source of genotypes from where they 
originated as well as the history of their quality data. For barley reference and spectral data analysis 150g per 
sample was taken after manually cleaned and graded. Then the samples were packed into plastic bag. Prior to 
reference samples chemical analysis, malt(after malted)were ground using a Laboratory Sample Mill3100 (Perten 
Instruments, Hagersten, Sweden) to pass through 0.5mm sieve for calibration  reference data. But for malt friability 
determination the malt sample was not grounded, because the friabilimeter machine itself grounds the sample for 
the ratio of friability measurement.Before malt quality traits analysis, the malt barley samples were malted 
according to Phoenix Automated Micro malting system (Phoenix Bios stems, Adelaide, Australia) designed to 
process 300g of 24 barley samples per batch [10]. After kilning the rootlets were removed from the malted 
samples by using mechanical malt cleaner that had been reconfigured to simultaneously process eight 250 g 
samples (Fraser Fabrications Pty Ltd, Malaga Western Australia). 
 

2.3. Wet Chemistry Analysis 
Malt quality traits of malt barley were chemically analyzed for reference data at Holeta Food science and 

Nutrition Laboratory, EIAR (Ethiopian Institute of Agricultural Research) in collaboration with VLB Institute in 
Berlin (Germany) for the traits mentioned below. But simple quality traits were analyzed at Holetta, EIAR cereal 
quality laboratory. Blank and known concentration sample were analyzed with the samples to control the analysis 
biases. The samples also duplicated to reduce the reproduced errors in each sample chemical analysis. 
 
2.3.1. MaltExtract Content  

Malt extract content was determined according to a small-scale version of the European Brewery Convention 
[11] Methods Manual, Section 4.9.1. Fine grind malt was extracted using a hot water mashing bath (SIEMENS 
Mashing Machine, Germany).For extraction, 50 g of finely ground malt was mixed with 200 mL of distilled water 
and mash at 45°C with continuous stirring. After 30 minutes of mashing, the temperature was increased by 
1°C/min until 70°C. As temperature reaches 70°C, there was added of 100 mL distilled water. After 1 hour, the 
mash was cooled to 30°C and adjusted to a volume of 515 mL or a weight of 450g. The extract was filtered using 



Agriculture and Food Sciences Research, 2020, 7(1): 38-45 

40 
© 2020 by the authors; licensee Asian Online Journal Publishing Group 

 

 

whatman 12cm filter paper into 500ml cylinder and specific gravity was measured at 20oC using a DMA5000 
density meter (Anton Paar GmbH, Graz, Austria). Therefore the following formula was used to put the end 
result.E = P(800+M)/(100-P); where, E= Extract content, P= Wort Density (oPlato), M= Malt Moisture content. 
 
2.3.2. Malt Total Protein  

The malt protein content was determined using kjeldhal method (Digester SBS 2000, Distillation Unit 
5000DL, FoodALYTGimbH, Germany) according to AOAC [12]. For analysis one gram ground sample of malt 
barley was measured and transferred into completely dry kjeldhal flask. Ten gram of kjeldhal tablet was added to 
the sample inside the flask. Twenty milliliter of 98% concentrated sulphuric acid was mixed with the sample. The 
sample digestion was started by connecting the kjeldhal flasks with the digestion rock. The digestion was 
completed when the brown color of the sample completely disappeared. After the digested sample was cooled, 100 
ml of distilled water and 80 ml of sodium hydroxide (32%) were added and distilled into 25 ml of excess boric acid 
containing 0.5 ml of screened indicator. The distillate was titrated with 0.1N hydrochloric acid to the methyl red 
end point. The protein was calculated by using this formula: CP%= [(T-B)*14*6.25)]/[W(100-MC)]; where 
CP=Crude Protein, T= Volume of HCl used in Titration, B= Blank used as control and W= Weight of sample 
taken for analysis. 
 
2.3.3. Malt Friability  

Malt grain samples were analyzed using a friability measuring machine (PfeufferFriabilimeter GmbH, 
Germany) which is used as pressure roller to grind the sample against a rotating screen. Low, medium and high 
friability malts were tested according to EBC method 4.15 [11]. 50g malt sample was run in the friabilimeter for 8 
min and the non-friable fraction was weighed to get the final result. 

 

2.3.4. Malt-Glucan  

The malt -glucan content was determined using the Megazyme kit method (Megazyme, Bray, 
Ireland)according to European Brewery Convention [11] Method 4.16.1.For the analysis 100 mg sample was 
suspended and hydrated in a buffer solution of pH 6.5 and incubated with purified lichenase enzyme and filtered. 

An aliquot of the filtrate was then hydrolyzed to complete with purified β-glucosidase. The D-glucose produced 
was assayed using a glucoseoxidase/peroxidase reagent. The final prepared aliquot was measured by 
spectrophotometer at absorbance 510 nm against reagent blank within one hour. Finally the beta-glucan was 

calculated using the formula;Β-glucan (%W/W) = ∆A * (F/W) * 27;Where, ∆A=Absorbance afterβ-glucosidase 

treatment (reaction) minus reaction blank absorbance, F=Factor for the conversion of absorbance values to μg of 
glucose, W= The calculated dry weigth of the sample analyzed in mg. 
 

2.3.5. Dry Matter Content  
Barley grain dry matter content was determined according to AOAC [12] international standard method from 

grain flour prepared using the above sample preparation method. 5g of barley flour was taken using a sensitive 
analytical balance and oven dried at 105OC temperature for 3 hours. After the dried sample was cooled in a 
desiccator, the final measurement was taken using the same analytical balance to get the result using the following 
known formula for moisture content. MC% = (Wi-Wf)/Wi*100; where Wi is initial weight, Wf is Final Weight; 
DM=(100-MC)%. 
 

2.3.6. Samples Spectral Data Acquisition 
Before scanning the samples, the spectroscopic performance of the device was checked using gold standard 

(1.038) and light trap standard (0.00065) provided by Bruker company of Germany. Then 60 barley samples were 
scanned twice for spectral data using near infrared spectrometer (Tango2017, Bruker Optics GmbH, Germany) 
which uses scan and rotating mode within 16 seconds, sample presentation with rotating accessory having 600 mL 
sample cuvette. It used32 amounts of scans between 867 –2535 nm wave length ranges with 16 cm-1 wave number 
band resolution. The NIR sensor used in this experiment is capable of recording spectra by diffusive reflectance 
measurements at the mentioned spectral wave length range. Spectra data figure-1 shown below was acquired from 
untreated barley grains before malting using a Bruker Tango software. It is the raw spectral data which needs to 
be treated statistically to correct spectral biases. The device recorded spectra in diffusive reflectance by using an 
integrating sphere. Measurements were done in duplicate for each sample, leading to a total number of 120 spectral 
data. 
 

 
Figure-1. Barley samples component matrix raw spectral data collected by NIR. 

 



Agriculture and Food Sciences Research, 2020, 7(1): 38-45 

41 
© 2020 by the authors; licensee Asian Online Journal Publishing Group 

 

 

2.4. Spectral Data Pre-Processing 
There were spectral biases and overlaps of wave length bands due to matrix effect, different particle size and 

spectroscopic condition. For this matter spectral data obtained according to figure-1 above must be pre-processed 
using the mathematical manipulation of the NIR machine as follows. The spectral data shown at figure-2 below 
was pre-processed using chemometrics system of FT-IR’s OPUS software version 7.5.1. Therefore, first Derivative 
plus Standard Normal Variate (1st D/ve+SNV) transformations, with 17 smoothing points were found to be the 
best method for the above spectral data treatment/pre-processing. The above Figure 1 raw spectral data was 
changed to the spectral data shown below in Figure 2. The shape and the resolution of the spectral data was 
completely changed as reflected below after preprocessed. So, the main objective of spectral data pre-processing 
was to remove biases, unwanted outliers and increase the resolution of overlapped wave length bands that could 
result errors in development of calibration model. 
 

 
Figure-2. Samples pre-processed spectral data. 

 

2.5. Developing Calibration and Validation 
The laboratory reference data was correlated with spectral data using the software OPUS version7.5.1 of 

Tango (Bruker, Optics GmbH, Germany). In total, 120 spectral data were used for calibration and validation. From 
120 spectral data, with approximately two thirds of the samples were used in the calibration set and one third in 
the validation set. Validation method used to check the performance of the calibration model was test set validation 
method. The calibration set was checked to cover the whole variation in terms of spectral data, traits and locations. 
The following numbers of factors were used for the traits of interest: protein content 8, extract content 8, friability 
9, ß-Glucan 10 and dry mater 10.The calibration model result was evaluated by using OPUS software statistical 
systems (chemometrics) [13].  
 

2.6. Calibration Model Statistical Evaluation 
Statistical tools were used to evaluate the efficiency of NIR calibrations and various terms are important in 

understanding the performance of a calibration model as described by Williams [14]. This includes statistics of 
calibration as well as statistics of validation. For developing calibration model a Partial Least Square (PLS) 
regression was used which was performed using OPUS software. 

The Coefficient of Determination (R2), Root Mean Square Error Estimation (RMSEE), Root Mean Square 
Error of Prediction (RMSEP),Ratio of performance to Deviation (RPD), Standard Deviation (SD) and Standard 
Error of Prediction) good parameters to evaluate the model performance depending on the reference data and 
spectra. The calibration was automatically tested by test set validation method. The final calibration was 
determined from an optimization routine of OPUS after the removal of the outliers. During the optimization step, 
various frequency regions and also spectral pre-treatments was systemically tested to determine the optimal 
calibration that was directly stored in the device memory for future quantitative analysis as used by Krapf, et al. 
[13].  

 

3. Results and Discussion 
Calibration model development highly depends on selecting a set of good calibration samples data obtained by 

wet chemistry analysis method. The set of calibration samples used in this study also contained the range of 
chemical and physical variations for which calibration model was applied. Accordingly, the calibration experiment 
was established using a mathematical relationship between the NIR spectrum and physical/chemical properties 
determined by reference methods. Similarly, comparing the chemical reference data variability and model accuracy 
is vital point. 
 

3.1. Calibration Model 
NIRS calibration models were developed for five barley traits determining quality for malt and food use. The 

model performance was assessed by the following parameters: coefficient of determination of the calibration (R²c) 
and validation set (R²v), standard errors of prediction (SEP) as well as root mean square error of prediction 
(RMSEP). The RPDv value indicates the suitability of the calibration for the prediction. With a higher RPDv value 
the calibration will more likely be able to predict the right sample values. According to the generally used 
classification in agriculture based on RPDv-values and R2

c-values, the calibration for protein can be regarded as 
successful calibration  [15]. 



Agriculture and Food Sciences Research, 2020, 7(1): 38-45 

42 
© 2020 by the authors; licensee Asian Online Journal Publishing Group 

 

 

The calibration and validation statistical description was very important to see the feasibility of each traits data 
to build calibration model. Table 1 implies the general features of descriptive statistics of traits to detect the right 
data variability for developing strong calibration model. Good data variability was observed in all traits except for 
DM as shown in Table 1 below. Therefore, expected calibration model is good for all traits if could not be strong 
model for DM only i.e. not implying the good fitness between reference data and spectral data, it is all about to 
capture the range of traits value obtained from different barley samples.  
 

Table-1. Descriptive statistics for calibration and validation.  

  Calibration Data Validation Data 

Traits Units Mean Range SD NC Mean Range SD NV 

Extract % 78.1 73-83.5 2.93 120 78.1 69.9-84 3.12 60 
Protein % 10.4 7-14.1 1.88 120 10.5 6.8-13.3 2.08 60 

Friability % 61.0 29-97 18.96 120 59 20-98 22.13 60 

β-glucan Mg/L 685.0 158-1000 279.5 120 699 50-1000 358.6 60 

DM % 91.34 90.6-92.3 0.4 120 91.28 90.8-91.9 0.26 60 
Note: DM=Dry Matter; SD=Standard Deviation; NC=No of Spectral Data Used for Calibration; NV=No of Spectral Data Used for Validation. 

 
The following barley traits analyzed chemically using wet chemistry methods were assigned for calibration 

model to fit the true value with predicted spectral value using PLS statistical regression. Spectral data was pre-
treated using appropriate statistical tool to remove outliers and to check the status of the chemistry and spectral 
data for developing calibration model as planned.  
 

3.2. Extract Content Model 
Extract content prediction delivered good calibrations as in Table 2 for whole grain samples (R2

c= 0.956; 
RPDC= 4.54 with variable samples). The model major parameters R2 and RPD reflect acceptable for most 
applications and screening purposes. Results from this study was comparable and much better than that of previous 
researchers who developed promising calibrations for predicting the extract of whole grain (R2

c =0.78 - 0.85) [16] 
and ground barley (R2

c = 0.77 -0.96) [17]. Because this property is highly influenced by the malting process since 
enzyme activity during malting influences the malt extract which limits the accuracy of any NIR prediction based 
on not malted barley. This is why different calibration model performance is reported by different scientists, even if 
the accuracy of reference sample analysis data is very important. Additionally, the following Figure 3 expressed the 
calibration curve obtained from chemistry/reference data and NIR predicted spectrum data of extract content.  
 

 
Figure-3. Barley extract content calibration curve. 

 

3.3. Protein Model 
Similarly calibration model results were obtained for whole barley grain samples protein content with (R2

c= 
0.97; RPD= 5.7 of variable samples)as in Table 2. The model is more acceptable than the models for other traits 
which could be usable in most applications, quality assurances and quality control. This prediction of nitrogen 
content from whole grain barley is well established in the literature and the results from this study compared well 
with those of previous reports for whole grain barley with R2

c = 0.94 [18] and R2
c = 0.95 [19]. Because NIR 

prediction is more effective in predicting biochemical properties than physical properties according to my 
observation from different literatures report. In addition to model parameter information, the following Figure 4 
expressed the calibration curve obtained from reference data and NIR predicted spectrum data of protein content. 

 
Figure-4. Barley protein content calibration curve. 

 



Agriculture and Food Sciences Research, 2020, 7(1): 38-45 

43 
© 2020 by the authors; licensee Asian Online Journal Publishing Group 

 

 

3.3.1. Friability Model 
Friability prediction performance model shown in Table 2 of this study was (R2

c=0.95; RPD=4.36), which was 
excellent like the protein model which could be used in most applications including in quality assurances, but in our 
case it was trustful model to identify excellent barley friability for malt factories at breeding final stage. Almost 
similar performed model was also reported in the literature with parameters value (R2

c=0.91; RPD=3.33). 
Therefore, the friability calibration model was not as such challenging like in dry matter and beta-glucan as 
observed from similar model reported by different authors [15]. Additionally, Figure 5 expressed the calibration 
curve obtained from chemistry/reference data and NIR predicted spectrum data of friability content. This curve 
was built after exaggerated outliers were removed from the spectrum data manually as well as by the OPUS 
system.  
 

 
Figure-5. Barley friability calibration curve. 

 

3.3.2. β-Glucan Model 
Beta-glucan prediction model was successful in this study for whole grain barley samples having (R2

c=0.90; 
RPD=3.18) as in result Table 2. According to Williams [14] this type of model performance is usable with caution 
specially for screening purposes such like in early stage breeding lines. But a similar study have been reported in 
the literature with much lower (R2

c=0.25)as compared to this study by Black and Panozzo [16]. On the other hand 
[20] reported (R2

c=0.61) and he referred that the poor distribution of reference values in the sample range may be 
the reason for the poor results in Beta-glucan model of barley grain that were obtained in his study. But in this 
study the good modifying malting process, excellent sample variability selection and moderate accurate chemistry 
analysis made the model better as compared to the reported literature. Additionally, the Figure 6 expressed the 

calibration curve obtained from chemistry/reference data and NIR predicted spectrum data of β-glucan content. 
 

 
Figure-6. Barley beta-Glucan calibration curve. 

 

3.3.3. Dry Matter Model 
Moisture predictions model from the same whole barley grain samples were good as in Table 2, but was only 

acceptable for some screening purposes, because (R2
c = 0.86; RPDc=2.69) was less as compared to the 

recommended range by Williams [14]. Results of moisture content from this study are not comparable well to 
literature reports, this is due to the smaller sample moisture content ranges (90.6-92.3 % DM) utilized compared to 
those used by previous researchers. Similar problem was also observed in literatures with the small sample range 
(78.4 - 83.4% DM) with in reference values obtained as studied by Roux [20]. The range of samples needs to be 
expanded in order to obtain acceptable calibrations model. Additionally, the Figure 7 expressed the calibration 
curve obtained from chemistry/reference data and NIR predicted spectrum data of dry matter content. 
 



Agriculture and Food Sciences Research, 2020, 7(1): 38-45 

44 
© 2020 by the authors; licensee Asian Online Journal Publishing Group 

 

 

 
Figure-7. Moisture content calibration curve. 

 
Generally, calibration model for the three traits extract content, protein content and friability were promising 

model for value predicting in some applications while beta-glucan and DM was delivered not trusted model as 
observed in the result Table 1 below. But calibration model alone does not enough to judge traits model to be 
applicable for targeted purpose. It is very important to see the validation parameters result side by side with the 
calibration parameters result as clearly shown in Table 2 for all selected traits in this study. 
 

Table-2. Calibration and validation statistical parameters results. 

 Calibration Statistical Parameters Validation Statistical Parameters 

Traits R2 SEC RMSEE RPD slope R2
V SEP RMSEP RPD Bias 

Extract 95.62 0.65 0.67 4.54 0.95 0.81 1.37 1.34 2.28 0.062 
Protein 96.93 0.33 0.34 5.7 0.97 0.93 0.55 0.54 3.78 -0.03 

Friability 94.75 4.35 4.5 4.36 0.95 0.59 14.10 13.8 1.57 0.32 

β-glucan 90.11 87.90 91 3.18 0.90 0.49 254.33 249 1.41 -22.6 

DM 86.23 0.15 0.153 2.69 0.86 0.51 0.18 0.29 1.43 -0.001 
Note: SD=standard Deviation of reference data; R2 =Coefficient of determination; SEC=Standard Error of Calibration; Standard Error of Prediction; 
RMSEE=Root Mean Square Error of Estimation; RPD=Ratio of Performance to Deviation. 

 

3.4. Traits Validation Model 
Model evaluation was performed by test set validation; because test set validation is a more independent 

validation method than cross-sectional validation as Williams [14]. In this method, the set of calibration samples is 
divided into a calibration set and validation set. According to this study from 60 samples,20 samples spectral data 
and chemistry data was set for validation. Using this process, the models were validated and checked for their 
prediction capacities. The principle was predicting quantitative value using the model from the validation set 
spectral data and comparing the predicted value against the chemistry data set for validation. 

Validation parameters for each barley traits were acceptable as in Table 2 for Extract content, protein content 
and friability having(R2

V= 0.81& RPD=2.28), (R2
V=0.93 &RPDV=3.78) and (R2

V=0.59& RPDV=1.57) values 
respectively. These models could be applied for quality control and screening purposes in breeding programs or 
any other applications. But as previously observed in calibration model above validation result for beta-glucan 
(R2

V= 0.49 with RPDV = 1.41) and DM (R2
V= 0.51 with RPDV = 1.43) was not reflected good model even though 

could be used in some rough screening purposes. This kind of model was not widely usable for prediction as Black 
and Panozzo [16] reported unless for simple rough screening purpose. Other parameters like SEP\RMSEP and 
Bias with lower value reflected that the model was well performed. These validation parameters showed very 
convincing value that the user could believe the model to use for prediction of protein, extract and friability in 
general. Similarly, a study by [18] delivered excellent prediction models for these traits that could be used in most 
applications. This should be due to the nature of the sample and less modifying malting process. 
 

4. Conclusion 
Barley is the fourth most important cereal crop worldwide. Also, it is a crop of ancient origin in Ethiopia and is 

considered as a center of diversity for barley, because of the presence of great diversity in ecology. Most of the 
Ethiopian barley is consumed as food at home. At the same time using barley for malt production establishes new 
value-added chains from which Ethiopian small holders can benefit significantly. The main objective of this study 
was enabling the breeding programs to select the appropriate genotypes easily by developing calibration model 
using near infrared spectroscopy. Depending on the objective Samples wet chemistry for barley (Extract, protein, 
friability, beta-glucan and moisture content) were analyzed using the international official methods described in 
materials and method part. Likewise, calibration models were developed for all above mentioned traits using NIR 
OPUS software statistical parameters. These wet chemistry and calibration model results demonstrated a realistic 
approach to predict quality traits of Ethiopian barley protein content, extract content, friability, ß-glucan and dry 
matter by applying NIRS.  

Therefore, this prediction model will enable the selection of appropriate food and malt barley genotypes for 
future end goals. Since NIRS is fast and cost-efficient the barley breeding programs can increase the intensity of 
selecting superior candidate lines for variety verification. Also with regard to broadening calibration performance 
Ethiopian barley landraces were included in the samples to capture the whole Ethiopian barley variability.  



Agriculture and Food Sciences Research, 2020, 7(1): 38-45 

45 
© 2020 by the authors; licensee Asian Online Journal Publishing Group 

 

 

5. Recommendation 
Malt quality traits that could have direct or indirect correlation with the identified traits should be included for 

model development to reduce matrix effect. For some of the traits like dry matter and beta-glucan further study is 
very important to distinguish between sample variability range and wet chemistry analysis inaccuracy which could 
contribute to results less trusted model observed in most studies. Also, to develop most excellent prediction 
method for barley, excellent sample variability, accurate wet chemistry data and having good instrument condition 
is very important. 
 

References 
[1] S. Marwat, M. Hashimi, and K. Khan, "Barley (Hordeum vulgare L.) A prophetic food mentioned in Ahadith and its ethnobotanical 

importance," American-Eurasian Journal of Agricultural & Environmental Sciences, vol. 12, pp. 835-41, 2012. 
[2] FAO, "Food balance sheets. Faostat. Rome. Retrieved from: http://faostat3.fao.org/download/fb/fbs/e," 2014. 
[3] S. E. Ullrich, Genetics and breeding of barley quality attributes. In: Barley Science: Recentadvances from molecular biology to agronomy of yield 

and quality (edited by G.A. Slafer, J.L. Molina-Cano, R. Savin, J.L. Araus& I. Romagosa). Binghamton, New York: The Haworth Press, 
2002. 

[4] S. Kreisz, Malting. In: Handbook of brewing: Processes, Technology, Markets (edited by H.M. Eßlinger). Weinheim: Wiley-VCH 
VerlagGmBH& Co. KGaA, 2009. 

[5] D. E. Briggs, Hough. JS, Stevens, R. & Young, TW,‘. Malting and brewing science'. London: Chapman and Hall, 1981. 
[6] T. Woodcock, G. Downey, and C. P. O'Donnell, "Better quality food and beverages: The role of near infrared spectroscopy," 

Journal of Near Infrared Spectroscopy, vol. 16, pp. 1-29, 2008. Available at: https://doi.org/10.1255/jnirs.758. 
[7] C. Pasquini, "Near infrared spectroscopy: Fundamentals, practical aspects and analytical applications," Journal of the Brazilian 

Chemical Society, vol. 14, pp. 198-219, 2003. Available at: https://doi.org/10.1590/s0103-50532003000200006. 
[8] H. Huang, H. Yu, H. Xu, and Y. Ying, "Near infrared spectroscopy for on/in-line monitoring of quality in foods and beverages: A 

review," Journal of Food Engineering, vol. 87, pp. 303-313, 2008. Available at: https://doi.org/10.1016/j.jfoodeng.2007.12.022. 
[9] W. Wang and J. Paliwal, "Near-infrared spectroscopy and imaging in food quality and safety," Sensing and Instrumentation for Food 

Quality and Safety, vol. 1, pp. 193-207, 2007. Available at: https://doi.org/10.1007/s11694-007-9022-0. 
[10] M. Nilsen and J. Panozzo, In: Leaping ahead with near infrared spectroscopy. Ed by Batten, G.D., Flinn, P.C., Welsh, L.A. and Blakeney, 

A.B. Australia: NIR Spectroscopy Group, 1995. 
[11] European Brewery Convention, Analytica EBC, 5th ed. Nurnberg, Germany: Verlag Hans Carl, 1998. 
[12] AOAC, Official methods of analysis, 17th ed. USA: Association of Official Analytical, 2000. 
[13] L. C. Krapf, A. Gronauer, U. Schmidhalter, and H. Heuwinkel, "Near infrared spectroscopy calibrations for the estimation of 

process parameters of anaerobic digestion of energy crops and livestock residues," Journal of Near Infrared Spectroscopy, vol. 19, pp. 
479-493, 2011. Available at: https://doi.org/10.1255/jnirs.960. 

[14] P. C. Williams, Implementation of near-infrared technology. In: Near-Infrared Technology inthe Agricultural and Food Industries (edited by 
P.C. Williams & K.H. Norris). USA: St. Paul, Minnesota, 2001. 

[15] G. Fox, M. G. Borgognone, P. Flinn, and D. Poulsen, "Genetic and environmental analysis of NIR feed quality predictions on 
genotypes of barley (Hordeum vulgare L.)," Field Crops Research, vol. 120, pp. 380-386, 2011. Available at: 
https://doi.org/10.1016/j.fcr.2010.11.007. 

[16] C. Black and J. F. Panozzo, "Utilising Near infrared spectroscopy for predicting malting quality in whole grain barley and whole 
grain malt," in Proceedings of the 10th Australian BarleyTechnical Symposium, Canberra, ACT, Australia, 2001. 

[17] S. Tragoonrung, P. Hayes, and S. Broich, "Near-infrared reflectance estimates of grain protein and malt extract in hill and row plot 
evaluations of spring malting barley," Canadian Journal of Plant Science, vol. 70, pp. 71-78, 1990. Available at: 
https://doi.org/10.4141/cjps90-008. 

[18] M. Edney, J. Morgan, P. Williams, and L. Campbell, "Analysis of feed barley by near infrared reflectance technology," Journal of 
Near Infrared Spectroscopy, vol. 2, pp. 33-41, 1994. Available at: https://doi.org/10.1255/jnirs.29. 

[19] M. Sohn, D. S. Himmelsbach, F. E. Barton, C. A. Griffey, W. Brooks, and K. B. Hicks, "Near-infrared analysis of whole kernel 
barley: Comparison of three spectrometers," Applied Spectroscopy, vol. 62, pp. 427-432, 2008. Available at: 
https://doi.org/10.1366/000370208784046768. 

[20] E. Roux, "Near infrared (NIR) spectroscopy for selection of malting barley in South African breeding programmes," Doctoral 
Dissertation, Stellenbosch: University of Stellenbosch, 2011. 

 

 

 

 

 

  

 

 

 

 

 

 

 

 
Asian Online Journal Publishing Group is not responsible or answerable for any loss, damage or liability, etc. caused in relation to/arising out of the use of the content. 
Any queries should be directed to the corresponding author of the article. 
 

http://faostat3.fao.org/download/fb/fbs/e,

