The quantitative analysis of samples led to the identification of honeys with a high quantity of pollen and fungal elements and others very poor in these elements. Pollen content varied between 1274 and 130,832 pollen grains per gram of honey, with a standard deviation of 25,585.

The qualitative pollen analysis identified 90 different pollen types in the samples. Table 1 lists the most commonly occurring pollens. This table includes 16 pollen types from ten botanical families, along with their corresponding maximum values, the standard deviation, and the range and the number of honeys in which the various pollen types reached a determined percentage level.

Castanea sativa, Cytisus type and Rubus were present in 100% of the samples, whereas Erica was present in 97.7% and Eucalyptus in 86%. These pollens were the principal types found in the honeys from north-western Spain. Echium, Quercus and Trifolium pollens were also frequently found. Pollens from Salix, Crataegus monogyna type, Campanula type, Conium maculatum types, Frangula alnus, Lotus type, Lithodora and Myosotis comprised more than 3% of pollen spectra and might also be important in honeys. Fifty pollen types were present only in a few honeys (less than 10%) and were included in other pollen type groups.

Regarding the botanical classification of honeys, 46 samples contained pollens from a variety of sources and were classified as polyfloral. Another 25 samples contained more than 45% Rubus pollens and were classified as blackberry monoflorals. Honeys containing more than 70% Castanea sativa pollen were classified as chestnut honeys (9 samples). The remaining 6 samples were Eucalyptus monofloral honeys with a percentage higher than 70%. Pollens from Erica plants were common, but none of the samples contained enough Erica pollen to be classified as heather monofloral.

Further microscopic analysis identified additional elements in the sediments from samples. Table 2 shows a descriptive analysis of principal fungal elements sorted by frequency. Cladosporium conidia were the best represented elements (occurring in 97.7% of samples), with a mean of 325 spores per gram of honey. Spores from Myxomycetes and Penicillium were also very abundant, with mean content of 135 and 151 spores/g, respectively. More than 40% of the samples contained these elements. The second-most common element was Metschnikowia yeast, present in 79.1% of the honeys studied. These cells have appeared in high content in some samples, with a maximum value of 46,217 cells per gram of honey.

Fungal spores produced by plant pathogens, such as Leptosphaeria, Stemphylium, Urediniospores, Alternaria, Pleospora and Botrytis, were also counted. Of these, Leptosphaeria was the most common (more than 50% of samples contain its spores), with a mean of 42 spores/g and a maximum of 613 spores/g. The next most common spores found were Stemphylium and Urediniospores (present in more than 30% of samples), with means of 19 spores/g and 25 spores/g, respectively. Other elements from plant pathogens were uncommon (occurring in less than 15% of the samples), with mean values of 6, 2 and 1 spores/g for Alternaria, Pleospora and Botrytis, respectively.

Basidiospores, which include spores from Basidiomycota other than yeasts, were present in 32.6% of samples with a maximum of 679 spores/g. A high number of spores from ferns (Pterydophyta) were present in two samples, probably because the bees collected them as a protein source. In relation to the honeydew index, the highest value was 0.63 and the lowest was 0.00, with a standard deviation of 0.09.

Table 3 shows the descriptive analysis for physicochemical parameters. The honeys had a low level of hydroxymethylfurfural (HMF), indicative of fresh honeys. The mean value was 0.3 mg/100 g; in fact, only three samples had more than 1 mg/100 g of HMF. Diastase varied between 6.1 Shade and 31.9 Shade, and invertase varied between 4.3 IN and 35.9 IN. Electrical conductivity and pH were two parameters widely used to distinguish between nectar and honeydew honeys. The pH varied between 3.5 and 5.0, values higher than 5.0 were not detected. The electrical conductivity varied between a minimum value of 0.224 mS/cm and a maximum value of 1.168 mS/cm. The humidity content had a mean value of 17.3 ± 0.8%. Colour varied between 39 and 150 mm, with a mean value of 95 ± 27 mm. Light amber colour (34–85 mm) accounted for 38.3% of honeys; amber colour (85–114 mm) for 41.9%; and dark amber colour (more than 114 mm) for 19.8%. The studied honeys were representative of the honeys produced in this geographical area, in which medium-high coloured honeys with medium or high electrical conductivity predominate.

The most abundant mineral in the honeys was potassium, with a mean value of 138.7 ± 62.8 mg/100 g. The phosphorus content varied between 31.5 mg/100 g and 3.0 mg/100 g, with a mean value of 8.8 mg/100 g. The average amount of magnesium was 7.8 ± 7.0 mg/100 g, and the calcium level was 9.0 ± 2.8 mg/100 g. Other mineral elements were present in low quantities.

A linear rank correlation analysis was used to examine the relationships between the variables studied. We used a non-parametric method, in which a linear correlation coefficient was calculated using Spearman’s linear rank correlation analysis. In this procedure, the variables with the strongest relationship to the provenance of the honey were used.

Table 4 includes the value of the correlation coefficient and the significance (P < 0.05, P < 0.01, P < 0.001) of its relationship with the variables obtained by microscopic analysis, that could be related to the source of the honey. The presence of Metschnikowia cells had a positive correlation with Eucalyptus and Rubus pollen content (P < 0.05) and a negative correlation with Cytisus t. (P < 0.01) and Erica pollen content (P < 0.001). The first taxon produced blossom honeys, and the other two were numerous in dark honeys.

The group called HD spores included the levels of Alternaria, Leptosphaeria, Stemphylium, Botrytis, Pleospora and Urediniospores. These plant pathogen fungi constitute an important subgroup of the identified fungal elements. This group had a negative correlation with the Eucalyptus pollen content (P < 0.001) and a positive correlation with Rubus pollen (P < 0.05).

Metschnikowia cells had negative correlation coefficients with most of the physicochemical variables (P < 0.001) other than pH (no significance) and humidity (positive correlation P < 0.01) (Table 5).

The presence of other fungal elements in honeys correlated positively with all physicochemical variables except humidity. HD spores in particular had a positive correlation (P < 0.001) with all the physicochemical parameters except humidity, with which it was negatively correlated (P < 0.05).

Regarding the honeys’ pollen content, Eucalyptus pollen had similar correlations to those of nectar yeast. Pollens from Cytisus and Erica types had positive correlations (P < 0.01) with electrical conductivity, mineral content (P < 0.05, P < 0.01, respectively) and colour (P < 0.01, P < 0.001, respectively). Cytisus t. also had positive correlations (P < 0.05) with diastase content, and Erica had negative correlations with pH (P < 0.01). Castanea pollen and Rubus pollen had positive correlation with electrical conductivity (P <0.05) and diastase (P < 0.05), respectively.

The cluster analysis was performed with the main variables obtained by microscopy as independent variables. These were Metschnikowia cells, HD spores, Castanea pollen, Eucalyptus pollen, Rubus pollen, Cytisus pollen and Erica pollen. The groups (honeydew honeys or blossom honeys) were established according to the characteristics of the samples, as previously pointed out. The cluster (Figure 1) represents the groups established in the analysis. Sixty four samples belonged to the blossom honey group (polyfloral honeys, Eucalyptus honeys, Rubus honeys and Castanea honeys) and 22 to the honeydew honey group (including blend honeys and honeydew samples).

The groups established by the cluster analysis were compared using a t-test for the equality of the means. Levene’s test examined the homogeneity of the variances by performing a one-way ANOVA on the absolute deviation scores.

Table 6 shows the descriptive statistical analysis of each group. Blossom honeys (N group) had a mean electrical conductivity of 0.540 ± 0.2 mS/cm, lower than that of the HD group (mean: 0.830 ± 0.2 mS/cm). Colour was clearer for nectar honeys (85 ± 20 mm pfund) than for honeydew honeys (124 ± 21 mm pfund). Mineral content was generally lower in blossom honeys, particularly for potassium content (N group mean: 120.4 ± 50.7 mg/100 g vs. HD group mean: 192.1 ± 65.2 mg/100 g), as was phosphorus content (7.4 ± 3.2 mg/100 g vs. 13.2 ± 8.2 mg/100 g). Even the enzymatic activity was high in the nectar samples. The subindex of the variables indicates the significance levels between the groups. Some parameters led to differentiation between the groups. These were enzymatic activity, Metschnikowia cells and Cytisus pollen type with p < 0.05, electrical conductivity, humidity, colour, potassium, calcium, magnesium, phosphorus, total mineral content, Eucalyptus pollen, Erica pollen and HD spores content with p < 0.01.

Pollen and the different fungal elements were analysed, identified and counted with a modified version of the method described by Louveaux et al. [31]. The quantitative analysis was made by a volumetric method using two aliquots of 10 µL taken from 2 mL of sediment obtained by centrifuging 10 g of honey. The total number of pollen grains and fungal elements were counted by microscopy in each aliquot and were expressed as an average. Pollen spectra were constructed by counting a minimum of 800 pollen grains in two 100 µL aliquots of the sediment obtained previously. The amounts of fungal spores and yeasts were also calculated from a 100 µL aliquot, with the number of pollen grains on the slide as a reference. The results were expressed as the quantity of the different fungal elements per gram of honey. Also, the relation between the number of fungal spores and the number of pollen grains (HDE/P) named as the honeydew index (HDE) was calculated.

The principal parameters of honey quality were determined using the methods adopted by the International Honey Commission [39].

Hydroxymethylfurfural (HMF) content was determined by UV absorbance of HMF at 284 nm. To avoid interference from other components at this wavelength, we determined the difference between the absorbencies of a clear aqueous honey solution and the same solution with added bisulphite, as described by White [40].

Invertase activity was determined by spectrophotometry of 4-Nitrophenyl-alpha-D-glucopyranoside decomposition at 400 nm. The samples were incubated at 40 °C. The invertase activity was expressed as an invertase number (IN). Diastase activity was based on the rate of starch hydrolysis by diastase, present in a honey buffer solution. The endpoint of this reaction was determined by measuring samples of the mixture at different time intervals using a UV-VIS spectrophotometer at 660 nm. The time required to reach an absorbance of 0.235 was calculated using linear regression, and the results were expressed using a Gothe scale. The UV-VIS absorbance was measured with a spectrophotometer (Jenway 6505).

The percentage of moisture was determined by refractometry, using an Abbe 325 refractometer. The pH was measured with a pH meter (Crison micropH 2001) of honey dissolved in bidistilled water. The electrical conductivity was determined with a conductivity meter (Knick 913 C), and the results were expressed as mS/cm. All measurements were performed at 20 °C. The colour was measured with a Hanna Honey Colour C221 digital instrument, and the results were expressed in millimetres using the Pfund scale.

The quantitative determination of minerals, such as potassium, calcium, iron, magnesium, sodium, phosphorus, zinc and copper, in the honey was conducted with an atomic absorption spectrophotometer (Varian SpectrAA-220). Samples were digested in a microwave before their mineral content was evaluated.

Physicochemical variables, mineral content, the most frequently occurring pollen types, the yeast content and the fungal content were statistically compared using Spearman’s rank correlation analysis.

A cluster analysis was applied to differentiate the honeys according to their botanical origin (either blossom honeys or honeydew honeys). The variables used were principal pollen types, Metschnikowia cells and a group of spores of plant pathogens called HD spores. Finally, the two groups established by the cluster analysis were compared with an independent samples t-test, which determines the presence of statistically significant differences between-groups.

All statistical analyses were performed with the SPSS Statistic 17.0 software for Windows.