Quality authentication and geographical origin classification of honey of Amhara region, Ethiopia based on physicochemical parameters

1 Introduction

According to the Codex Alimentarius definition, honey is the natural sweet substance produced by honey bees, Apis mellifera, from the nectar of plants (blossoms) or from the secretions of living parts of plants (honeydew) or excretions of plant-sucking insects on the living parts of plants, which honey bees collect, transform by combining with specific substances of their own, deposit, dehydrate, store and leave in the honeycomb to ripen and mature (Alimentarius, 2001). The majority of its components are derived from the plants; while some are added by the bees and others are due to the maturation of the honey (Teklit and Frehiwot, 2016).

From the chemical point of view, the composition of honey principally depends on the plant species visited by the honey bees. Besides, environmental and climatic factors, the processing conditions during its formation, harvesting process, and storage conditions do have a significant role (Nayik and Nanda, 2016). In general, about 200 various natural substances are found in honey. Essentially, it is composed of carbohydrate (mainly glucose and fructose) that constitutes 80–85% followed by water which comprises 15–17% (Buba et al., 2013). It has also a wide range of minor constituents that were found at trace amounts like vitamins, inorganic minerals, enzymes, as well as many phenolic acids and flavonoids (Manyi-Loh et al., 2011; Soares et al., 2017).

Honey is popular and widely used nutrition, a raw material for traditional beverages, and many medicinal activities as well as ingredients or additives in the food industry. As a result of its wide use, consumption and relatively high price, illegal honey adulteration either in production or in geographical and/or botanical origins becomes a series problem. The former is related to the addition or removal of various substances during honey production while the latter is related to mislabeling for some economic interests (Moloudian et al., 2018). Due to such practices, the honey industry has given due attention to authentication practices throughout the world. The European Union(EU) council directive, the Codex Alimentarius and other international legislations do have honey regulation indicating the permissible limit of various parameters that helps to ensure quality and authenticity to protect consumers from product adulteration and fraud (Alimentarius, 2001; Council, 2002).

Various researchers justified that using physicochemical and organoleptic parameters such as moisture content, HMF-value, sugar composition, electrical conductivity, ash content and the like can be used for routine quality control and authenticity of honey (Ruoff and Bogdanov, 2004; Boussaid et al., 2018; Soares at al., 2017). Such physicochemical parameters were taken as good quality indicators by many researchers and used as screening tools to identify impaired honey. Analysis of sugar, organic acids and amino acids from a very fresh honey, investigation on electrical conductivity and HMF can be used again for quality control of honey by comparing with standards as reported by many researchers (Gebremariam and Brhane, 2014; Getachew et al., 2014; El Sohaimy et al., 2015). Of course modern spectral analysis like infrared and chromatographic analysis of the pollens, chemical compositions or additive ingredients were found to be applicable to test for quality and authenticity of the honey samples (Ruoff and Bogdanov, 2004; Soares et al., 2017).

Furthermore, the European Union (EU) council directive put forward that botanical and geographical origins of the product must be declared on the package label targeting guarantee product quality and authenticity for the benefit of consumers (Council, 2002; Karabagias et al., 2014).

Several approaches together with multivariate data analysis tools have been proposed to characterize the floral and geographical origin of honey (Soares et al., 2017). Multi-element analysis and isotope ratio (Fechner et al., 2020; Soares et al., 2017) volatile constitutes (Panseri et al., 2013), amino acids profile (Rebane et al., 2008), individual sugar profile (Soares et al., 2017), phenolic acids analysis (Tomás‐Barberán, et al., 2001), and pollen analysis (Tomás‐Barberán, et al., 2001; Soares et al., 2017) have been repeatedly proposed to trace out honey of a particular origin. To profile these constituents, the state of the art analytical techniques such as inductively coupled plasma- mass spectrometry/optical emission spectrometry, liquid chromatography-mass spectrometry/tandem mass spectrometry, gas chromatography-mass spectrometry/tandem mass spectrometry, and/or nuclear magnetic resonance spectroscopy are required (Soares et al., 2017). These analytical tools can provide accurate data, but they are expensive and hence could not be used for routine analysis in an ordinary lab. Consequently, a simple and easily accessible approaches based on physicochemical and sensory parameters together with statistical modeling have been proposed to classify honey of different geographical and botanical origins (Bogdanov, 1997; Piro et al., 2002; Ruoff and Bogdanov, 2004; Moloudian et al., 2018; Oroian et al., 2017).

However, to the best of our knowledge, such in-depth work related to authenticity of the quality and geographical origin classification was not reported on the honey of the Amhara region, Ethiopia. So the target of this work was to quantify the physicochemical parameters of the region’s honey and to see its quality by comparing it with the values of international standards and to classify the region’s honey based on suitable statistical models.

2 Materials and methods

2.1

2.1 Sample collection and the study area

Forty-seven honey samples were collected from seven different zones of Amhara region; the northern part of Ethiopia shown in Fig. 1. Amhara region is one of the major honey producing regions in the country. The honey samples were fresh as they were collected directly from the traditional local beekeepers in October, November and December 2019. After the honey samples were collected preliminary purification with sieves like separation of dead bees, wax sticks, particles of combs and other debris were done and then stored in a refrigerator at a temperature of 4 °C. The sampling areas or zones of the region were shown on the map as in Fig. 1. And the collected samples with their codes, geographic origin, color and climate were presented as in Table 1

Close

Fig. 1

Map of sampling administrative zones taken from the Google map and modified.

Export to PPT

Table 1 Description of the collected samples.

Sample Code	Zone	Climate	Visual Colour	Sample Code	Zone	Climate	Visual Colour
W1	North Wollo	Midland	Yellow	G1	West Gojjam	Midland	Amber
W2		Highland	White	G2		Midland	Amber
W3		Lowland	Amber	G3		Lowland	Amber
W4		Lowland	White	G4		Lowland	Yellow
W5		Midland	Amber	G5		Lowland	Amber
W6		Lowland	Yellow	G6		Lowland	Yellow
W7		Lowland	Yellow	G7		Lowland	Amber
W8	Wag Himera	Lowland	White	G8		Lowland	Yellow
W9		Lowland	Amber	G9	East Gojjam	Midland	Yellow
W10		Highland	White	G10		Midland	White
W11		Highland	White	G11		Midland	Yellow
W12		Midland	White	G12		Highland	White
W13	South Wollo	Lowland	Yellow	G13		Highland	White
W14		Midland	Amber	G14		Lowland	Yellow
W15		Midland	White	G15	Awi	Highland	Amber
R1	South Gondar	Lowland	Yellow	G16		Highland	White
R2		Lowland	Yellow	G17		Highland	Amber
R3		Lowland	Yellow	G18		Highland	Yellow
R4		Midland	Yellow	G19		Highland	Amber
R5		Midland	Yellow	G20		Highland	Amber
R6		Highland	White	G21		Highland	White
R7		Midland	Yellow	G22		Highland	Yellow
R8		Highland	Amber	G23		Highland	White
R9		Highland	White

Key: Samples Coded by; W = are from Wollo province, R = are from Gondar province, G = are from Gojjam province

3 Results and discussions

The physicochemical parameters values such as moisture content (MC), pH, free acidity (FA), total reducing sugars (TRS), apparent sucrose content (ASC), electrical conductivity (EC), ash content (AC), and Hydroxymethylfurfural (HMF) of the 47 honey samples analyzed in this study are presented in Table 2. In this study, the ANOVA result showed the existence of significant variations on the values of the physicochemical parameters among the 47 honey samples taken from different geographical origins at P < 0.05 (at 95% confidence interval).

Considering the moisture content, the value ranged from 14.20 g in a sample at the north Wollo zone to 19.74 g in a sample at the west Gojjam zone per 100 g of honey. The minimum and maximum mean values were obtained 16.06 ± 0.31 g from North Wollo zone and18.51 ± 0.15 g from Awi zone (mean ± SE) per 100 g of honey respectively where SE is the standard error of the mean. The maximum value of moisture content that quality honey should hold was given by the European Directive union, Codex Alimentarius Standard and Quality& Standards Authority of Ethiopia. These three organizations suggest that the moisture content of quality honey should be ≤ 21 g/100 g of honey sample (Alimentarius, 2001; Bogdanov et al., 1999; QSAE, 2005). The findings showed that the regions honey full fills the requirement and accordingly, it is less prone to fermentation and do have a longer shelf life. The moisture content is one of the most important characteristics, influencing many other properties of honey, like viscosity, crystallization, color, flavor, taste, specific gravity and solubility. High moisture content indicates a premature extraction or extraction under high humid conditions that could lead to granulation, fermentation, spoilage and loss of flavor ensuring the deterioration of honey quality (Gebremariam and Brhane, 2014; Kaur et al., 2016).

$\mathit{SE}=S/\sqrt{n}$ ; Where n is the number of data or samples; S is the standard deviation; SE is Standard Error of the mean, Max. is maximum value, Min is minimum value.

The findings on moisture content in this study are in agreement with the results reported on honey from other regions of Ethiopia (17.28 ± 0.28 to 20.05 ± 0.18 g/100 g) (Berhe et al., 2018). Comparable moisture content in honey was also reported from Romania (14.44 ± 0.02 to 19.89 ± 0.01%) (Oroian et al., 2017).

As the moisture content, pH value of the honey is a nice quality parameter. Naturally honey is acidic due to the presence of organic acids, and inorganic ions such as phosphate and chloride. Acidity may also be enhanced by the storage conditions and extraction process (Karabagias et al., 2014). In general, a low pH of honey inhibits the growth and proliferation of microorganisms that have a role in its stability, shelf-life and texture of the honey (Moloudian et al., 2018). The pH values of the individual sample in the present study ranged from 3.63 in a sample at Wag Himera zone to 4.65 in a sample at West Gojjam zone. Furthermore, the minimum mean value of pH was noted from Wag Himera zone, 3.85 ± 0.06, while the maximum mean value (4.23 ± 0.05) was from west Gojjam zone (mean ± SE). This mean value is within the limits of international standards from 3.2 to 4.5 as cited in (Tesfaye et al., 2016). The findings of this work are in agreement with the reported data on honey from other regions of Ethiopia, 3.52 – 4.52 (Berhe et al., 2018)and Tunisia, 3.45 – 4.63 (Boussaid et al., 2018).

The free acidity of the individual honey sample varied from 18.00 meq/kg at West Gojjam zone to 40.00 meq/kg of honey at four different zones as shown in Table 2. Also, the mean value of the free acidity ranged from 26.39 ± 1.80 meq/kg for samples from east Gojjam zone to 37.80 ± 1.87 (mean ± SE) for samples from Wag Himera zone. These findings are in the acceptable range stated at the international level by Codex Alimentarius standards,≤ 50 meq/kg of a honey sample as well as at the national level by Quality and Standards Authority of Ethiopia given to be ≤ 40 meq/kg of a honey sample(Alimentarius, 2001; QSAE, 2005). High acidity may give a sour taste to the honey and deteriorate its quality (Oroian et al., 2017; Tesfaye et al., 2016). On the other hand, free acidity values with moisture content are suggested to be excellent parameters for geographical origin classification by pattern recognition techniques (Moloudian et al., 2018).

The content of total reducing sugar (TRS) in the analyzed honey samples ranged from 56.50 to 82.64 g/100 g of sample. The lowest and the highest TRS values were noted in samples from the Awi zone and West Gojjam zone respectively. Considering the average values of TSR at each sampling zone, it was found that samples from the Awi zone had the lowest mean TRS value (61.45 ± 0.92 g/100 g), whereas samples from South Wollo and Wag Himera zones were found to have the highest mean value of TSR (71.41 ± 0.98 g/100 g). The internationally accepted minimum limit of TRS suggested by codex is ≥ 45 g/100 g honey (Alimentarius, 2001)and for European Union directive is ≥ 60 g/100 g of honey(Council, 2002). The mean result of TRS from each zone in the present study satisfies this minimum limit put by the organizations. Comparable results were reported in the honey of other regions of Ethiopia (Getachew et al., 2014; Gobessa et al., 2012).

The average sucrose content of the honey samples ranged from 2.96 ± 0.30 to 4.73 ± 0.09 g/100 g of honey (mean ± SE). These lowest and highest average sucrose content were found in honey samples from east Gojjam zone and south Wollo zone respectively. Whereas, in considering the individual honey samples, a 0.45 g lowest value from West Gojjam zone and 5.84 g highest value of sucrose per100g of honey at two zones as shown in Table 2 were noted. The sucrose level should not exceed 5 g/100 g of the honey sample according to the codex and European Union council directive of international regulatory standards (Alimentarius, 2001; Council, 2002). Though few of the samples at various study zones contain sucrose slightly above this permissible limit, still the mean value of sucrose content at each study zone satisfies the quality requirement given by these organizations. Analogous results were reported in Bale honey (other regions of Ethiopia) ranged from 3.01 to 7.62% (Tesfaye et al., 2016).

The electrical conductivity of honey is closely related to the concentration of mineral salts and it is a parameter showing great variation in their origins or sources. Blossom honey usually has lower mean values of electrical conductivity, in comparison to honeydew honey and it is a nice honey quality parameter (Gebremariam and Brhane, 2014). The minimum and maximum electrical conductivity values of the honey under investigation were found to be 22.50 μS/cm at Wag Himera zone and 85.10 μS/cm in a sample from Awi zone at 20 °C. And the average value was ranged from 33.37 ± 2.07 to 69.64 ± 3.00 μS/cm (mean ± SE) at 20 °C in Wag Himera and Awi zones respectively. These results were within the acceptable limit (<800 μS/cm) given by Codex Alimentarius (Alimentarius, 2001). A lower electrical conductivity result of honey ranged from 8.27 to 33.5 µs/cm was reported from other regions of Ethiopia (Nigussie et al., 2012). Other research from Saudi Arabia has shown much high electrical conductivity values ranging from 215.2 to 3138.8 µs/cm (Alqarni et al., 2014).

The average ash content of honey samples in the present study was ranged from 0.26 ± 0.03 to 0.60 ± 0.01 g/100 g of honey (mean ± SE) in East Gojjam and West Gojjam zones consecutively. The minimum and maximum values were found to be 0.02 g and 0.62 g per 100 g of honey at South Wollo and South Gondar zones respectively. Though few of the samples possess slightly higher ash contents, the mean values obtained here were in harmony with the international as well as nationally acceptable limits which are ≤ 0.6% by mass (Alimentarius, 2001; Council, 2002; QSAE, 2005). The ash found in honey expresses its richness in mineral content, used as an origin classification parameter and one quality criterion (Karabagias et al., 2014). Comparable findings were reported from Iran (0.041–0.562%) (Moloudian et al., 2018) and in Gambella, other regions of Ethiopia (0.28 to0.41 g/100 g) of honey (Berhe et al., 2018).

Hydroxymethylfurfural (HMF) is a fructose degradation sugar naturally present in all honey at traces or probably not at all at the initial time. Quantification of HMF in honey samples is a reflection of either its freshness or an excellent indicator of quality deterioration in honey samples based on the amount present (Gebremariam and Brhane, 2014; Moloudian et al., 2018). In general, the best quality of honey has a lower amount of HMF value. The findings of this research work (Table 2) showed that the range of HMF values obtained from each sample was 3.04 at the south Gondar zone to 22.87 mg/kg of honey at the Wag Himera zone. Whereas the average HMF values were noted from 3.88 ± 0.14 mg at South Gondar to 8.50 ± 1.39 mg from south Wollo zone (mean ± SE) per kg of honey.

The amount of HMF for quality honey according to codex Alimentarius Standard is ≤ 60 mg/kg honey(Alimentarius, 2001). While the European Directive Union and Quality & Standards Authority of Ethiopia both suggest being ≤ 40 mg/kg honey (Council, 2002; QSAE, 2005). Based on these findings, the region's honey is in excellent quality condition. Even the very low HMF content in the honey samples indicates that the honey samples produced and collected in the study area were extremely fresh and its freshness consequently tells us there is a good honey handling practice in the region.

In a Summary, the result obtained from the analysis of each physicochemical parameter of the 47 honey samples collected from the seven administrative zones of the Amhara region versus the international and national permitted values were presented in Table 3.

Where MC: moisture content, FA: free acidity, TRS: total reducing sugar, ASC: apparent sucrose content, EC: electrical conductivity, AC: ash content.

On the other hand, the findings of this study showed that white-colored honey possesses a comparatively higher mean value of apparent sucrose content and free acidity but lower in its ash content, electrical conductivity and hydroxymethylfurfural value. The amber type honey possesses higher electrical conductivity, ash content and lower values of free acidity, and apparent sucrose content.

The yellow-colored honey type enjoys almost medium values between white and amber-colored honey as depicted in Fig. 2. However, the moisture content and total reducing sugar were not significantly different (P > 0.05) in the three classes of honey, based on their color.

Close

Fig. 2

Average values (mean ± standard error) of the physicochemical parameters based on the color of the sample: Yellow, White, and Amber type honey. The means across a bar for the same parameter with different letters were significantly different (P < 0.05); while those indicated by the same letter are insignificant as determined by Ducan’s post hoc test. (MC: moisture content, FA: Free acidity, TRS: total reducing sugar, EC: Electrical conductivity, ASC: Apparent sucrose content, AC: ash content, and HMF: hydroxymethylfurfural).

Export to PPT

The mean data values obtained based on the climatic conditions of the sample area also showed that; honey from highland gets higher moisture content and free acidity but lower in its total reducing sugar and HMF values. Honey from low land acquires higher HMF value and lower moisture content. The midland honey does have moderate values in its moisture content and HMF values as demonstrated in Fig. 3. However, the apparent sucrose content and electrical conductivity values didn’t show any significant difference (P > 0.05) between the three honey categories.

Close

Fig. 3

Average values (mean ± standard error) of the physicochemical parameters based on the climate of the sampling area: Lowland, Midland, and Highland. The means across a bar for the same parameter with different letters were significantly different (P < 0.05); while those indicated by the same letter are insignificant as determined by Ducan’s post hoc test. (MC: moisture content, FA: Free acidity, TRS: total reducing sugar, EC: Electrical conductivity, ASC: Apparent sucrose content, AC: ash content, and HMF: hydroxymethylfurfural).

Export to PPT

In all cases, the variation of the physicochemical parameter values between the different sampling regions, climatic conditions and colour types might be due to the difference in soil composition and climatic conditions, the natural chemicals that the bee colony may add during its formation and the contamination with other strange materials during production, handling and storage processes (Aljohar et al., 2018; Teklit and Frehiwot, 2016).

4 Multivariate analysis

Correlations among variables were shown in Fig. 4. Some of the variables are negatively correlated like free acidity with conductivity while some others correlate positively like pH with electrical conductivity. And the correlation between some of the components is strong like in pH and free acidity while others do have less strong correlations. Besides the correlation, the significance of the mean values of each of the physicochemical parameters for the three provinces (Gojjam, Gondar, and Woll honey) was performed using one way ANOVA. According to the significance values obtained from the ANOVA test, the mean values of the physicochemical parameter (except for apparent sucrose content, ASC and ash content, AC) were found to be significant (p < 0.05). While those mean values of ASC and AC of the samples taken from the three provinces were not significant (P > 0.05). This information was helpful in constructing the LDA model.

Close

Fig. 4

Correlation matrix of honey samples variables (47 samples × 8 variable).

Export to PPT

Principal Component Analysis (PCA):- The results of the eight physicochemical parameters on the 47 honey samples were subjected to a principal component analysis to observe the leading eigenvectors of the correlation matrix that contain most of the variances. PCA was used to visualize sample trends and evaluate the discriminatory characteristics of the determined physicochemical parameters. Thus the PCA result showed us the first three principal components having eigenvalues>1 explains 32.14%, 21.18%, and 15.82% consecutively; which explains about 69.14% of the total variations.

The score plot for the first two components was shown as part of it as in Fig. 5. This tells us the PCA potentially reduces the dimensions and complexity of a data matrix with a loss of only a few of the information. The loading plot in Fig. 6 showed us the first component was most dominated by pH, free acidity and electrical conductivity. While the moisture content and total reducing sugar were mainly loaded on the second component and HMF was loaded on component 3.

Close

Fig. 5

Scores plot of the principal component analysis model constructed from the physicochemical parameter values of the honey sample.

Export to PPT

Close

Fig. 6

Loading plot projections of the physicochemical variables used in PCA.

Export to PPT

Both the score and loading plot on Fig. 5 and Fig. 6 showed us most of the Gojjam province honey samples positioned towards the positive of component 1 and the negative of component 2 were connected vastly to moisture content, pH and conductivity values. Most of the honey samples from Wollo province located at the left part of the score plot was associated with free acidity, total reducing sugar and apparent sucrose content while the honey samples from Gondar province placed themselves at the middle part of the score plot.

Besides this, from the loading plot, we can see a vivid correlation among each variable. For example, pH and free acidity do have inverse relation which is a known fact. The same variation can be observed between the sugar content and the moisture content. The other important information is that pH, free acidity, moisture content, total reducing sugar and electrical conductivity variables comprises most of the variations as predicted from the length of the loading plot projection. As obtained from the ANOVA test and confirmed from the short loading projections ASC and AC contributes less for the variations.

Linear discriminant analysis (LDA) was used to describe differences between groups and to exploit those differences in allocating (classifying) new observations to the groups. The former one is discrimination while the latter is prediction. The model discriminates 86.96% (20 of the samples) of the Gojjam province’s honey, 55.56% (5 of the samples) of the Gondar province’s honey and 86.00% (13 of the samples) of Wollo province’s honey Fig. 7. The model does have an average of 80.85% discrimination power.

Close

Fig. 7

Scatter plot of the linear discriminant analysis model based on the physicochemical profiles of the 47 honey sample.

Export to PPT

In the discrimination analysis model, Fig. 7, the Gojjam province honey samples were located below 0 in discriminant score 1 and between −1.5 and 1.9 in discriminant score 2. And the Gondar province honey is positioned above −0.7 in discriminant score 1 and below −0.3 in discriminant score 2 while the Wollo province honey placed themselves above 0 in discriminant score 1 and above −0.4 in discriminant score 2. These data can be used in allocating new observations in either of the groups. The LDA model was validated using cross-validation methods. In this method, the sample was randomly divided as a training data set that contains 39 samples (about 80%) of the sample and testing or validation data set which contains the rest 8 samples. The recognition power of the model which was obtained from the training set and found to be 84.62% and the prediction power of the model which was obtained from the validation set was 100.00% as shown in Fig. 8 A and B.

Close

Fig. 8

Recognition and prediction abilities of the LDA model tested in a cross validation method. A) Training data set B) Validation data set.

Export to PPT

Hierarchical cluster analysis (HCA) was applied to examine the natural groupings (or clusters) of observations in the dataset based on their dissimilarity measures. For HCA the dendrogram was employed as a graphical tool to show the clusters based on the similarity ratios.

Though dividing the dendrogram to determine the number of clusters is a subjective process, let us see the nature of the dendrogram by forming four clusters. The first cluster enclosed by a green rectangle contains mainly the Gojjam honey from Awi Zone. And the second cluster separated by a red rectangle is again the Gojjam honey mainly from east and west Gojjam zones (Fig. 9). Most of the honey from Wollo province forms a group at the end cluster enclosed by violet color rectangles while the Gondar province honey is interlinked with some of the honey samples from Wollo and set at the 3rd group enclosed by a blue rectangle (Fig. 9). Some of the honey samples were not separately clustered and hence intermixed in the dendrogram. The reason might be due to the similarity in the environmental conditions (altitude, climate, and probably their soil type and composition).

Close

Fig. 9

Dendrogram from Agglomerative Hierarchical Cluster Analysis of 47 honey samples using their physicochemical parameters.

Export to PPT

5 Conclusion

This study gives more attention to the quantitative determination of eight physicochemical variables to relate their values with the quality of the honey and developing chemometric models for geographical origin classification that helps for discrimination purpose on the honey samples harvested at various administrative zones of Amhara region, Ethiopia. The result confirmed that all the 47 honey samples were found to have an excellent quality based on the national as well as international honey quality standards given by the Codex Alimentarius, Council Directives of the European Union and quality and standards authority of Ethiopia (QSAE). The PCA resulted in a reduction of dimensions and complexity of a data matrix and by LDA it was possible to discriminate the samples 80.85% on average. On the validation of the LDA model its recognition power and prediction capacity was found to be 84.62% and 100% respectively.