*2.5. Data Analysis*

Sensorial data was analyzed using the XLSTAT Sensory tool by Addinsoft (Paris, France). This package let to obtain the characterization of the samples considering the attributes perceived by the 8 tasters. The procedure for sample characterization and panel evaluation was used. Significant statistical differences were set as *p*-value < 0.05. The radar charts for monofloral samples were done using the factorial values obtained in the characterization of products.

The package Analyzing data of XLSTAT was used as an exploratory statistical tool to reduce dimensionality of the multivariate data and to visualize them graphically, with minimal loss of information. PCA analysis was applied to identify groups of samples according to their botanical origin. This multivariate analysis allowed to summarize the information includes in the variables studied into a small number of principal components or factors providing a simplified interpretation of data variance through mathematical methods.

### **3. Results and Discussion**

### *3.1. Sensorial Profile of Samples*

Organoleptic properties are the first attributes that consumer can observed. These comprise visual properties as state, color, smell and aroma perceptions and taste. The descriptors considered in this work, to describe honey samples, had different discriminant power (Figure 1). The most useful were visual color, the different families of smell (animal, chemical, floral, fruity and vegetal), the persistence of smell, aroma perceptions (mainly fruity, vegetal, animal and floral families) and the saltiness for taste descriptors. The rest of the descriptors have been poorly detected so presented non-significant values.

**Figure 1.** Sensorial descriptors used and their discriminant power.

All the samples were crystallized when the sensorial analysis was performed. Hence the scale used for color descriptors was from white to brown. The clearest samples were classified like straw color (two samples), 13 had gold color, five orange color and 10 brown color. The size of crystals was very fine or fine in 14 samples, medium in 15 samples while only one sample had large crystals. The tasters described the intensity of the smell as low being the predominant the vegetal in 21 samples, fruity in three and floral in two samples; however, secondary perceptions were mainly fruity, floral and in six samples, animal. The persistence of the odor was generally low or extremely low. Regarding the taste, besides the sweetness, tasters detected a degree of sourness and saltiness in most of the samples while bitterness was less common. When the aroma was considered, the most common perceptions were fruity, floral and in lesser extent vegetal. The most usual descriptors for vegetal smell or aroma were leaves, wood, mint, resin or cinnamon, for fruity were fresh fruit, tropical fruit, peach, dry grapes or fig, for floral were rose, lavender and violet and for animal leather and wax (except two samples in which tasters made associations with urine). Astringency was detected in a few samples. Finally, it should be mentioned that 15 samples had an intense smell to smoke (degraded family), that could be caused during the honey harvest. To protect quality beekeepers should improve honey managemen<sup>t</sup> techniques.

### *3.2. Pollen Spectra and Main Botanical Origin of Samples*

Ninety-six different pollen types were identified in the pollen spectrum of the honey samples including 77 nectariferous taxa and 19 taxa which do not produce nectar. Samples present a grea<sup>t</sup> diversity of pollen types, the mean value was 26, but 12 samples had more than 30 different pollen types. The botanical families Fabaceae, Asteraceae, Apiaceae and Lamiaceae provide the 36.0% of the pollen types, standing out the diversity of Fabaceae plants which were 15.3% of identified types. Within this family, *Genista* and *Hedysarum* pollen types were the most frequent pollen grains. The first was present in 96.7% of the samples and the second one in 83.3% of them.

*Genista* was dominant pollen in three samples and the maximum value reached was 73.9% whereas *Hedysarum* was dominant pollen in two samples and had a maximum value of 63.2%. For *Asteraceae* plants, *Galactites* t. was found in 70.0% but the values in the pollen spectra were lower than 7%, while *Aster* t., identified in 50.0% of honeys, had a maximum value of 39.7%. The most frequent Lamiaceae pollen was *Stachys* t., found in 70% of samples with a maximum value of 9.7%. Other common pollen type was *Thymus* t. (46.7% of samples) but values were always below 3%. Finally, Apiaceae plants mainly as *Foeniculum vulgare* t. and *Pimpinella anisum* t. were found in 73.3% and 50.0% honeys, respectively (Table 3).


**Table 3.** Frequency classes of the main pollen types.


**Table 3.** *Cont.*

1 Max.: maximum value reached by the pollen type in samples, D: number of samples in which the pollen type is dominant (percentage ≥ 45%), A: number of samples where the pollen type is between 45% and 15%, R: number of samples where the pollen type is between 15% and 3%, I: number of samples where the pollen type is between 3% and 1%, present: number of samples where the pollen type was identified. t.: pollen type common for different plant genera.

> Despite the frequency of the mentioned botanical families in the honey samples, some other plants highlighted in the pollen spectra of honeys. The first was *Erica* genus, represented by the species *E. arborea* and *E. multiflora*. This pollen type was found in 70.0% of samples and had a maximum value of 72.8%, being dominant pollen in four samples and secondary pollen in seven samples. In addition, Myrtaceae is well represented in the area due to the presence of *Eucalyptus*, used to reforestation, and *Myrtus* that grows in Mediterranean area commonly below 600 m, both are present as secondary pollen in samples. *Pistacia* is other representative Mediterranean taxa, was present in 73.3% of honeys frequently with values under 15% but was secondary pollen in two samples. The area of study had an important rainy period that contributes to facilitate the introduction of some forest species in mountain areas as *Castanea*. The genus was only found in two samples, but its high bee value merit is to be considered. In the case of Rosaceae plants, the stand out, *Rubus*, presented in 70% of samples as being in one dominant pollen. Another mention should be made to the presence of *Asparagus* pollen, with a value of 21.0% in one sample.

> The honey samples had medium pollen content in quantitative terms. Values ranged from 1117 to 23,196 grains/g, with an average of 6710 grains/g. Most of the samples (66.7%) were classified in class II of Maurizio and 23.3% in class III. Only 10% of samples had low pollen content (Class I). This is according to the use of centrifugation to extract honeys from the combs.

### *3.3. Quality Evaluation of the Honey Samples*

Overall, most of the honey samples showed acceptable quality parameters that were in accordance with the international legislation on honey quality [18].

The evaluation of the degree of freshness gave good results for most of the samples. HMF values were below the limit of 40 mg/kg, having a mean value of 9.0 mg/kg. The lowest value was 1.3 mg/kg and the highest value was 31.8 mg/kg. Regarding diastase content, samples had extremely low values with a mean value of 11.3 DI. The minimum diastase level was 4.8 DI and the maximum was 29.0 DI. Ten honey samples had a diastase activity below 8 DI (the minimum legal limit in international standards without considerations about HMF content).

One important parameter for honey stability is water content. It varied from 16.4% to 19.8%, with a mean value of 18.1%. Most of the samples were collected during July and August and only one in April, but there were no differences in water content in relation to the date of harvest. Electrical conductivity was in general medium with a mean value of 0.60 mS/cm, a minimum of 0.29 mS/cm and a maximum value of 1.35 mS/cm. For pH, the values oscillated from 3.5 to 4.4 with a mean value of 3.9. The values of electrical conductivity correspond mainly to blossom honeys; however, six samples presented values over 0.80 mS/cm. Regarding to free acidity, all samples had values below the maximum permitted in European legislation which is 50 meq/kg [18], ranging from 14.1 meq/kg to 46.5 meq/kg with a mean value of 32.4 meq/kg. According to Pfund scale, the color of samples varied from extra light amber (37 mm), to dark amber (135 mm) with a mean value of 77 mm. In addition to the differentiation between blossom honeys and honeydew honeys, the EC of honey is strongly related to its organic acids, ash content and proteins [19]. Therefore, the higher the content of the latter, the higher the obtained conductivity. A high content of organic acid and salts increases the free acidity present in honey.

### 3.3.1. Mineral Content

Considering the average value of samples, the most abundant minerals were K followed by P, Na, Ca and Mg. The average value of K was 135.1 mg/100 g varying from a minimum of 43.1 mg/100 g to a maximum of 394.5 mg/100 g. P varied from 17.6 mg/100 g to 36.6 mg/100 g and had an average of 25.8 mg/100 g and Na, averaged 11.0 mg/100 g and ranged from 3.5 to 24.9 mg/100 g. Ca and Mg, had similar mean values of 7.5 mg/100 g and 5.5 mg/100 g, respectively, whereas Fe, Mn, Cu, Zn, were found with the lowest values ranging from the maximum value of 1.2 mg/100 g for Fe in some samples to values lower than 0.2 mg/100 g. Cd and Pb were always under the limit of detection. Normally, darker honeys had higher mineral content so heather honeys, chestnut honeys and honeydew honeys have the main values [20]. It has been proved by [21] that the botanical origin of honey has an impact on its elemental composition.

### 3.3.2. Total Phenolic and Flavonoid Content

The phenol content (TPC) and flavonoid content (TFC) are other interested parameters in honeys. TPC values ranged from 41.8 mg GAE/100 g to 128.3 mg GAE/100 g, with a mean value of 70.6 mg GAE/100 g, while TFC were in accordance with TPC and ranged from 2.3 mg QE/100 to 9.7 mg QE/100 g, with a mean value of 4.9 mg QE/100 g.

These results were included within the interval found by [22] on honey samples from the same region and close to those reported by [23] on honeys from Algeria. Furthermore, TPC values were included in the range reported by [24] on Italian honey. Regarding TFC it is possible to find a grea<sup>t</sup> variation in values depending on the methodology used, but these values are higher than other reported from Moroccan honey [25] and Czech honey [26].

Both parameters are related with botanical origin of honey and hence with color, mineral content and electrical conductivity [20]. Previous studies have demonstrated a strong relationship between the phenolic profile of different honey types and their antioxidant capacities [20,27,28]. Therefore, further study of biological activities of honey produced in Babors Kabylia, such as the antioxidant power of honey samples, would be very interesting and necessary to show the correlation between the botanical origin and the content of phenolic compounds.

The most critical values for the quality of these samples correspond to water content and diastase index. All the analyzed samples presented a water content below the limit established by international legislation. However, values were higher than those reported for honeys from semiarid areas of Algeria [29]. Indeed, honeys produced in dry regions should have lower moisture content than honeys produced in humid regions. Similar values were reported by [7,30] on honeys produced in a nearby region, as well as those produced in the Mediterranean coast [6], coinciding with the humid area of Algeria. In any cases, the studied samples had relatively high-water content that should be considered due

to their relationship with stability and conservation of honey. Furthermore, high quality honeys should contain low HMF content and high Diastase index. Both parameters are thermosensitive and increase with aging and prolonged heating of honey [31]. Diastase content was frequently lower than the stablished limit in international legislation; however, HMF content does not exceed the limit. Similar values to those found in these sample was reported by [31] for Algerian honeys but higher values were also found [7]. It is essential to remark that water content and diastase index are dependent on several factors, including botanical and geographical origin, environmental and seasonal conditions, the degree of maturity and the beekeepers' handling during the honey harvest [32]. Studies on the common practices of honey managemen<sup>t</sup> by beekeepers are necessary to determine the degree of influence on honey quality. In any case, the lack of professionalism combined with the quasi-total absence of training sessions related to beekeeping, hive maintenance, honey resources in the region and harvesting management, in addition to the propagation of Varroa disease and the absence of a legislative framework for quality recognition, can lead to problems linked to the quality of Algerian honey. In this context, to promote competitiveness with imported honey in the conventional market, it is recommended to promote knowledge of the characteristics of local honey and the participation of all stakeholders who can take actions to maintain the quality of the product.

### *3.4. Multivariate Analysis Applied to the Interpretation of Samples Characteristics and Typification of the Honey Samples*

Data on the sensorial analysis, the main pollen types in the pollen spectra of samples and some physico-chemical parameters were used for PCA analysis. In total, 34 standardized variables were introduced to create the covariance matrix. The five first factors explained 92.0% of the variance of the data, and the first two 77.6%. The first principal component (PC1) represented 62.3% of the variance and had the main correlation coefficients with the parameters of visual color (0.843), Pfund color (0.968), TPC (0.976), TFC (0.959), and *Erica* (0.856). The second one (15.2% of the variance) had the highest positive correlation coefficient with *Genista* (0.936) and negative with *Hedysarum* (−0.733). Some of the variables are highly correlated and appeared together in the biplot (Figure 2a). At the right, sensorial perceptions such as vegetal odor and aroma, saltiness, bitterness and sourness taste and persistence of taste and smell, pollen types such as *Erica* or *Pistacia* and electrical conductivity, color, TPC, TFC and PH. At the left, are animal smell and sweetness taste together the pollen types *Eucalyptus*, *Myrtus*, *Rubus*, *Aster* and *Hedysarum*.

In the biplot of the two first components, the cases are dispersed but some similar samples appeared close (Figure 2b). At the right, it can be seen a group of samples with reference H (heather samples) and one Hd (honeydew sample). Up are the samples with the highest content in *Genista* pollen type and down, at the left, are the samples with the highest content of *Hedysarum*, *Aster* and the sample with *Rubus* as dominant pollen. In the center, there are a group of samples classified as polyfloral honeys (P).

Analyzing the results, main taxa for honey production in this hotspot for biodiversity were *Erica*, *Hedysarum*, *Genista* plants and in lesser extent *Myrtus*, *Eucalyptus* and some Asteraceae plants. The importance of these plants for honey production was mentioned before by [4,33,34].


**Figure 2.** Principal Component Analysis (PCA). (**a**) Loading biplot of the variables included in the analysis, (**b**) Score biplot of the samples regarding component 1 and 2; P: Polyfloral, H: Heather.

(**b**)

The shrubs occupy a large surface in deforested areas and slopes of the mountains. These Mediterranean maquis has a grea<sup>t</sup> interest for apiculture, being the most representative species *Erica arborea* developing mainly on nutrient-poor acidic soils and *Erica multiflora* growing in calcareous soils. Both have been found in samples and give recognized and high valued unifloral honeys. Other plants such as *Ceratonia siliqua* L., *Pistacia*, *Calicotome*, *Spartium junceum* L., and *Genista*, and some Lamiaceae like *Thymus* or *Lavandula*, *Phillyrea latifolia* L. or different species of *Cistus* and *Quercus* form plant communities with *Erica*, so their pollen appears in honeys samples. It is worth mentioning the controversy on the apicultural value of Genisteae plants; while some species are considered that do not produce nectar (such as *Calicotome spinosa*), others are considered good nectariferous source for honeys such as *Retama sphaerocarpa* [35] or *Spartocytisus supranubius* (L.f.) Christ ex G.

Kunkel [36]. In the case of some *Genista* species, *Chamaecytisus* and *Retama* nectar secretion was reported [37]. Furthermore, through the pollen spectra of samples, the grea<sup>t</sup> diversity in pollen types can be observed, and further, the high number of pollen types by sample in accordance with biodiversity of the area. In this sense, melissopalynology is a key analysis with which to identify geographical origin of honeys and to determine melliferous species in the area where honey has been collected [38].

Mainly, the apiaries were in altitudes lower than 600 m above sea level and near to villages. In these areas, shrubs and spontaneous plants are most common; therefore, they predominate in the pollen spectra of honeys. Spontaneous species have high melliferous yield; among them, standing out are the Fabaceae plants such as *Hedysarum*, which is represented in the area for some species [39], many species of *Trifolium*, *Lotus* and other as *Securigera atlantica* Boiss and Reut., *Scorpiurus muricatus* L., *Tetragonolobus purpureus* Moench, *Astragalus echinatus* Murray. and *Medicago polymorpha* L. The importance of Fabaceae for honey production in the region was mentioned previously, either to produce monofloral honeys and/or to participate importantly in the production of polyfloral honeys [4,33,40–42]. Their importance also appears in studies carried out on honeys from the Mediterranean regions of Algeria [2,7,34,43].

Other important herbaceous plants for honey yield were Lamiaceae and Asteraceae; both are the most visited after Fabaceae [44]. The best represented pollen type in samples belonging to the family Lamiaceae was *Stachys*, which includes plants like *Stachys ocymastrum* or *Phlomis bovei* Noë and many other plants of this family; these are probably exclusives of the area. Something similar occurs with Asteraceae plants; the best represented pollen type was *Aster* t. with a pollen grain morphology similar for plants such as *Aster*, *Helichrysum stoechas* (L.) Moench, *Dittrichia viscosa* (L.) Greuter or *Phagnalon saxatile* (L.) *Cass.*, frequently occurring plants in the area. Other pollen types of Asteraceae were identified in samples in lesser extents, but due to the possible contribution of these plants to honey production and the fact that they normally appear underrepresented in the pollen spectra of honeys [35], a deep investigation about vegetation and plant distribution in the area could help in the identification of honeybee resources.

### *3.5. Characteristics of the Different Honey Types*

The interpretation of the data allowed us to classify 12 samples, such as unifloral six heather honeys, two sulla (*Hedysarum*) honeys, two *Genista* honeys, one blackberry (*Rubus*) honey, one Asteraceae honey), one like honeydew honey and the rest, as polyfloral.

Table 4 shows the mean and standard deviation values for physicochemical and main pollen parameters. Polyfloral honeys showed the highest variation between samples in most of the variables as correspond to the heterogeneity of the group. A grea<sup>t</sup> variation in pollen content within polyfloral samples was noted, ranging from class I, II and III, although some heather honeys, *Hedysarum*, Asteraceae and honeydew were found to be in class III. Regarding color, polyfloral honeys were mainly light amber, as found in *Genista*, *Hedysarum*, Asteraceae and *Rubus* honeys. However, honeydew and heather honeys were the darkest honeys. The honeys presented similar water content that ranged from 16.8 to 19.6%. However, in terms of electrical conductivity and pH, honeydew honey presented the highest values. At the other extremes were *Genista* honey with the lowest electrical conductivity and *Rubus* honey with the lowest pH.



Although *Genista* honeys had the lowest free acidity, all honey types had similar values. The same occurred with the HMF content. Honeydew honey had the highest average value of diastase index, while the lowest level was observed in heather honeys. This difference could be explained by the fact that the concentration of diastase depends on several factors, including handling conditions, and could not be a reliable indicator of the honey's origin. It could therefore be explained by the fact that the honeys may have been heated or improperly stored. The TPC and TFC also varied depending on the type of honey. Honeydew honey followed by heather honeys had the higher values of these compounds. Finally, the mineral content was higher in honeydew honey. Secondly, the high content of K, Ca, Mg and Fe in heather honeys were noteworthy. However, followed by Honeydew honey, the P content was higher in Asteraceae and *Hedysarum* honeys. *Rubus* honey type showed a light amber color but also a low TPC and TFC, although this type of honey showed a wide variation in color and varied from extra light amber to dark amber [45]. It is therefore necessary to study more *Rubus* honeys from this region to verify this finding.

The sensorial profile of the different honey types can be seen in Figure 3. The six samples of heather honey presented a common pattern with dark color (brown), when crystallized, predominant vegetal smell (leaves and resin), some floral features and fruity (dry and tropical fruit). The persistence of the smell was low whereas the persistence of the taste was medium, and samples presented saltiness, bitterness and sourness in similar intensity. The aroma was mainly fruity and vegetal. The *Genista* samples had similar pattern in smell and aroma but different color, one had straw color and the other orange color (both were crystallized). The smell was mainly floral and in lesser extent fruity and aroma was mainly fruity. The persistence of smell was low, and the persistence of the taste was medium, the intensity of the sweetness was high and presented some saltiness and sourness notes. Sulla honeys were the most different regarding their sensorial profile, one was darker than the other and the smell and the aroma of one of them was animal (leather) while the other had floral and fruity smell. The persistence of the smell and the taste were similar in both cases. Asteraceae samples presented orange color, fruity smell and aroma with low intensity and persistence for smell, aroma and taste. Regarding the blackberry honey, highlighted the floral and fruity smell and the fruity aroma as well as the predominance of sweetness in taste. Finally, the honeydew sample had brown color, vegetal odor, fruity aroma and the highest persistence of smell.

The published scientific information about sensory profile of honey is scarce [12,35,46,47] and no information is available about sensorial properties of honey produced in Algerian country. The heather honeys were the most known and have similar sensorial profile to those previously described [35,48]. Regarding the sulla honeys were mainly described with floral smell but these samples were different in the smell having one of them fruity smell and the other animal smell, both with very low intensity [49]. Finally, the sensorial properties of Genista honeys are poorly described in literature, but its sensorial properties are near to other honey obtained from Fabaceae plants as *Spartocytisus supranubius* [36]. Other honey types are poorly known regarding their sensorial properties.

This is the first study focuses on the sensory properties of honey and their relationships with the botanical origin and the physicochemical properties of honeys from Babors Kabylia Region. Considering that this area is a hotspot for biodiversity and have a high potential for honey production, further studies could contribute to increase knowledge in the characteristics of the honey, the types that can be obtained and the most relevant plant species for honey yield and definitely contribute to the valorization of local beekeeping.

**Figure 3.** Chart diagrams for sensorial perceptions of monofloral honeys.
