Impaginato 329 Adv. Hort. Sci., 2023 37(3): 329­341 DOI: 10.36253/ahsc­15064 Characterization of Italian honeys: integrating volatile and physico­ chemical data C. Taiti 1, G. Guardigli 1, S. Babbini 1, E. Marone 2, E. Masi 1, D. Comparini 1 (*), S. Mancuso 1, 3 1 Dipartimento di Scienze delle Produzioni Agroalimentari e dell’Ambiente, Università degli Studi di Firenze, Viale delle Idee, 30, 50019 Sesto Fiorentino (FI), Italy. 2 Facoltà di Bioscienze e Tecnologie Agro‐Alimentari e Ambientali, Università degli Studi di Teramo, Campus Coste S. Agostino, Via R. Balzarini, 1, 64100 Teramo, Italy. 3 Fondazione per il Futuro delle Città, Via Boccaccio, 50, 50133 Firenze, Italy. Key words: Honey, honey characterization, honey origin, honey properties, honey volatiles, monofloral honey, Proton Transfer Reaction Time­ of­Flight Mass Spectrometer. Abstract: This article focuses on the comprehensive characterization of Italian honeys using various physico­chemical analyses and their volatile organic com­ pounds (VOCs) fingerprint obtained through the PTR­ToF­MS technology. Honey characteristics, including pH, electrical conductivity, moisture content, hydroxymethylfurfural (HMF), and sugar content, were analyzed to assess their quality and origin. Honey samples from different flowers, including acacia, chestnut, citrus, linden, and multifloral, were collected and investigated. Furthermore, a few aged honeys were collected and analyzed and compared with the fresh ones. Physico­chemical analysis revealed that chestnut honey is characterized by high pH and EC values. Acacia honey has a higher fructose con­ tent, while aging appears to influence HMF levels, a vital indicator of honey quality, with aged samples exhibiting significant increases in HMF content. The VOC profiles have been found to vary among different honey types, suggesting that VOCs could be used as indicators of honey origin. Multivariate statistical analyses, such as partial least squares discriminant analysis (PLS­DA), have been applied to the VOCs data to differentiate honey types based on their volatile profiles. Acacia honey exhibited different physicochemical parameters but on the contrary, in the VOCs analysis, it displayed similarities with the lin­ den honey due to their shared low emissions of volatile compounds. Citrus honey had similar chemical parameters to linden and multifloral honeys, but its distinctive VOCs emission allowed for a more accurate identification. In conclu­ sion, the analysis performed with the PTR­ToF­MS was successful in obtaining specific volatile fingerprints of those samples and was effective for improving the characterization of honeys. 1. Introduction Honey is a natural product known and used by humans since antiquity (*) Corresponding author: dr.comparini@gmail.com Citation: TAITI C., GUARDIGLI G., BABBINI S., MARONE E., MASI E., COMPARINI D., MANCUSO S., 2023 ­ Characterization of Italian honeys: integrating volatile and physico‐chemical data. ­ Adv. Hort. Sci., 37(3): 329­341. Copyright: © 2023 Taiti C., Guardigli G., Babbini S., Marone E., Masi E., Comparini D., Mancuso S. This is an open access, peer reviewed article published by Firenze University Press (http://www.fupress.net/index.php/ahs/) and distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. Data Availability Statement: All relevant data are within the paper and its Supporting Information files. Competing Interests: The authors declare no competing interests. Received for publication 24 August 2023 Accepted for publication 20 October 2023 AHS Advances in Horticultural Science https://doi.org/10.36253/ahsc-15064 http://www.fupress.net/index.php/ahs/ http://creativecommons.org/licenses/by/4.0/ http://creativecommons.org/licenses/by/4.0/ http://creativecommons.org/licenses/by/4.0/ Adv. Hort. Sci., 2023 37(3): 329­341 330 (Nikhat and Fazil, 2022). The Italian legislation, trans­ posing Directive 2001/110/EC, defines honey as “the natural sweet substance produced by Apis mellifera bees from the nectar of plants or from secretions of living parts of plants or excretions of plant sucking insects on the living parts of plants, which the bees collect, transform by combining with specific sub­ stances of their own, deposit, dehydrate, store and leave in honeycombs to ripen and mature”. Honey characteristics, such as flavour and physico­chemical properties, can vary substantially depending on botanical and geographical origin (Zhou et al., 2002; Warui et al., 2019). The Italian legislation (D. lgs. 21/05/2004, n. 179) has established thresholds and values for physico­chemical criteria, including mois­ ture, electrical conductivity, hydroxymethylfurfural (HMF), sugar, and others, to evaluate the marketabil­ ity and quality of honey, which were added to the aromatic profile of honey. Nonetheless, there are roughly 320 distinct types of honey available on the market, which can be grouped into monofloral and multifloral varieties (Vîjan et al., 2023). In Italy, there is a rich assortment of honeys, and this diversity is the result of the unique combination of regional pro­ duction, climate conditions, and a multitude of floral sources (Castiglioni et al., 2017). Monofloral honey is obtained from bees that have mainly visited a unique botanical species, these honeys are particularly valu­ able on the market (Schuhfried et al., 2016). As reported by ISMEA, the cost of multifloral honey dif­ fers from the cost of monofloral honey (ISMEA, 2023). However, European legislation does not speci­ fy the properties of monofloral honey, so countries like Italy imposed a national regulation with a mini­ mum percentage of pollen required for the identifica­ tion as monofloral, which varied from floral origin depending on the pollen production, position and flower structure of each botanical species (Tedesco et al., 2022). On the other hand, multifloral honey is produced from several types of flowers, and its char­ acteristics and properties can differ greatly depend­ ing on the visited flowers and the geographical origin. Melissopalynological analysis is the official method for identifying the botanical and geographical origin of honey (Aronne and De Micco, 2010). However, this analysis is time­consuming and cannot be applied to filtered honey. Moreover, the execution requires palynological competence, which is a limiting factor (Mureșan et al., 2022). In previous studies, PTR­ToF­ MS has been used for the categorization of honey types based on their aromatic profiles, such as the monofloral classification (Kuś and van Ruth, 2015; Schuhfried et al., 2016) and for the discrimination of their botanical origin (Ballabio et al., 2018). Thus, the primary objective of this study was to comprehensively characterize Italian honeys by employing a combination of volatile compound analysis, alongside conventional physico­chemical analyses. By integrating volatile profiling using PRT TOF­MS with established analytical techniques, the aim was to determine whether this analysis could serve as an additional, complementary or substitute method for discriminating different botanical origins of Italian honey. 2. Materials and Methods Sample collection Honey samples were gathered in 2022 from May to August directly from beekeepers from different natural geographical macro­areas (districts) of Italy to have a variety of sources that include region, province, altitude, and botanical origins. A total of 84 samples of honey were collected, 78 of these were obtained in 2022, and 6 were collected between 2020 and 2021 (aged samples). Each sample was stored in the dark in a cool and dry place. The collec­ tion focused mostly on Italian artisan­produced honey as reported in Table 1. In addition, to achieve even more powerful results, we collected 12 Italian commercial samples. The types of honey were 49 multifloral (of which 6 commercial), 16 acacia (of which 2 commercial), 11 chestnut, five citrus (of which 4 commercial) and three lindens. Physico‐chemical analysis All the physico­chemical analysis were performed according to the guidelines of the Italian regulation DM 25/07/2003 GU number 185 (Gazzetta Ufficiale, 2003). Determination of pH. To assess the pH, 10 g of sample was thoroughly mixed in 40 ml ultrapure dis­ tilled water (dilution 1:5) from a Millipore Milli­Q lab water system. The resulting solution was measured using a PHM 210 Standard pH Meter (MeterLab, Radiometer Copenhagen), which was previously cali­ brated with standard pH 4 and pH 7 solutions. Electrical conductivity. The EC of honey was obtained from the same diluted solution used to assess the pH. The measurement was done using a conductometer (Conductimeter GLP 31 CRISON) cali­ Taiti et al. ‐ Characterization of Italian honeys 331 brated with appropriate standard solutions. The results were expressed in mS/cm. The maximum EC for honey is 0.8 mS/cm according to Italian law (Directive 2001/110/EC), while for the honeydew, multifloral/mixed, and chestnut honey the EC values must be greater than 0.8 mS/cm. Moisture content. The water content of the honey samples was determined with a handheld refrac­ tometer (HHTEC) with automatic temperature com­ pensation. The samples were measured as­is, and the results are expressed as moisture content percent. The legal threshold for selling honey is 20%, but in competitions for premium/quality honeys, the limit is usually lowered to 18%. Hydroxymethylfurfural (HMF) and furfural (F) quantification by HPLC. The HMF and F were quanti­ fied following the HPLC method, which had been pre­ viously described in other studies with a few adjust­ ments in accordance with Italian legislation guide­ lines (Fallico et al., 2004; Truzzi et al., 2012). Briefly, 5 g of honey was diluted with ultrapure distilled water (1:5) and mixed. Then, within 12h, samples have been filtered on a 0.45 μm syringe filter and 20 μl were injected into the HPLC system (Azura, Knauer, Berlin, Germany) coupled to a UV detector (Analytical UV Flow Cell detector UVD 2.1S, Knauer). The chro­ matographic column was Eurospher II 100­5 C18 150 x 4 mm, and the analysis conditions were: isocratic mobile phase, water­methanol 90:10 v/v; flow rate 0.6 mL/min; column temperature 30°C. The detector wavelength was fixed at 285nm, the identification of HMF and F was done by comparing the retention time of standard solution, and the quantification was done using a calibration curve specific for each mole­ cule (Fig. 1 A). The calibration curve for HMF was made with five solutions at different concentrations (0.0005, 0.005, 0.01, 0.05, 0.1 mg/ml), while the F calibration curve was 0.0006, 0.001, 0.002, 0.006, 0.01 mg/ml. According to the law, the results were expressed in mg/kg, and the legal limit for HMF in commercial honey is 40 mg/kg. Sugars determination Brix determination. Brix degrees of the honey samples was measured with the same refractometer of moisture content measurement. Brix degrees rep­ resent the percentage of sugar content in honey by weight, with 1 Brix degree equivalent to 1 g of sucrose in 100 g of solution (Geană et al., 2020). Sugar quantification by HPLC. HPLC coupled to a refractive index detector was used for the qualitative Botanical Origin Region Province Production source Harvest year No. of samples Acacia Tuscany Firenze Beekeeper 2022 5 Acacia Tuscany Livorno Beekeeper 2022 1 Acacia Tuscany Arezzo Beekeeper 2022 1 Acacia Abruzzo Pescara Beekeeper 2022 1 Acacia Tuscany ­ Commercial 2022 1 Acacia Italy ­ Commercial 2022 1 Acacia Lombardy Cremona Beekeeper 2022 1 Acacia Abruzzo Teramo Beekeeper 2022 1 Acacia Tuscany Prato Beekeeper 2022 1 Acacia Piedmont Torino Beekeeper 2022 1 Acacia Tuscany Pisa Beekeeper 2022 1 Acacia Emilia­ Romagna Forlì Beekeeper 2022 1 Chestnut Tuscany Firenze Beekeeper 2022 4 Chestnut Tuscany Livorno Beekeeper 2022 1 Chestnut Tuscany Arezzo Beekeeper 2022 1 Chestnut Lombardy Cremona Beekeeper 2022 1 Chestnut Piedmont Torino Beekeeper 2022 1 Chestnut Tuscany Pisa Beekeeper 2022 1 Chestnut Emilia­ Forlì Beekeeper 2022 1 Chestnut Campania Salerno Beekeeper 2022 1 Citrus Calabria ­ Commercial 2021 1 Citrus Italy ­ Commercial 2021 1 Citrus Italy ­ Commercial 2022 2 Citrus Sicily Ragusa Beekeeper 2022 1 Linden Tuscany Firenze Beekeeper 2022 1 Linden Lombardy Cremona Beekeeper 2022 1 Linden Emilia­ Forlì Beekeeper 2022 1 Multifloral Tuscany Firenze Beekeeper 2020 1 Multifloral Tuscany Firenze Beekeeper 2021 1 Multifloral Tuscany Firenze Beekeeper 2022 21 Multifloral Tuscany Prato Beekeeper 2021 1 Multifloral Tuscany Livorno Beekeeper 2022 1 Multifloral Tuscany Arezzo Beekeeper 2022 3 Multifloral Abruzzo Pescara Beekeeper 2022 1 Multifloral Italy ­ Commercial 2021 1 Multifloral Italy ­ Commercial 2022 5 Multifloral Lombardy Cremona Beekeeper 2022 4 Multifloral Abruzzo Teramo Beekeeper 2022 1 Multifloral Sicily Ragusa Beekeeper 2022 3 Multifloral Piedmont Torino Beekeeper 2022 1 Multifloral Umbria Todi Beekeeper 2022 1 Multifloral Tuscany Pisa Beekeeper 2022 1 Multifloral Lazio Roma Beekeeper 2022 2 Multifloral Campania Salerno Beekeeper 2022 1 Total 84 Table 1 ­ Description of the traits of the samples analyzed, con­ sidering the different botanical origin, geographical area, and year of production Adv. Hort. Sci., 2023 37(3): 329­341 332 and quantitative analysis of sugars (AZURA RID 2.1L, Knauer, Berlin, Germany). The chromatographic col­ umn was Eurospher II 100­3 NH2 150 x 4 mm, employing a mobile phase 80:20 of acetonitrile­water and an isocratic flow rate of 1.5 ml/min at 35°C. Honey samples were prepared by placing 0.5 g in 10 mL of H2O (1:20), mixing for 12­24h, then filtering and diluting 1:1 using the same solution as the mobile phase. Calibration curves were prepared using fructose, glucose, sucrose, and maltose stan­ dards for quantification, and retention times were used for identification (Fig. 1 B). Six distinct solutions, each with a different concentration, were employed to construct the calibration curve. The concentra­ tions used were 0, 2.5, 3.75, 5, 7.5, 10, and 15 mg/ml. Defect identification by sensory analysis Before to test, each samples were homogenized by mixing with a glass rod, filtered and left until com­ pletely clear, after which they were subjected to organoleptic analysis (consistency, color, smell and taste) according to the national standard SR 784­ 3:2009 (Council European Union, 2001). Particular attention was directed towards identifying any potential defects in the honey samples, with a specif­ ic focus on the detection of fermentation. To confirm the conformance of honeys and eventually exclude samples with imperfections, the odour, colour, taste, and texture of honey were assessed. All the samples were found to be conforming and free from defects based on that assessment. PTR‐ToF‐MS measurements and data analysis Using a PTR­ToF­MS 8000 (Ionicon Analytik GmbH, Innsbruck, Austria) with H3O+ as the reagent ion and over the mass range of m/z 20­250, volatile finger­ prints of 84 samples were acquired. The benefits of the PTR­MS technology are fully and completely described in a previous study (Blake et al., 2009). Volatile headspace from each sample was ana­ lyzed follow the setup previously proposed by Schuhfried et al. (2016) with some modification. In short, 5 g of honey (±0.1 g) were placed into a 250 ml glass jar with two Teflon septa on the cap’s opposing sides for the VOCs analysis. Then, each jar has been sealed and fluxed with clean air for 60 seconds before the incubation time, in order to remove all the VOCs accumulated during the sample preparation. Subsequently, the samples have been incubated at 37°C for 30 min in order to allow VOCs to fill the head­space. Finally, the volatile compounds were analyzed using the PTR­ToF­MS in its standard config­ uration. The zero air­generator (Peak Scientific Instruments) supplied clean air at a flow rate of 0.5 lpm (lpm = liter per minute) to the entry of the sam­ pling device during all analyses, and the same flow rate was set for the PTR­MS inlet flow. To prevent the systematic memory effect, clean air was fluxed for five minutes in the tool apparatus between mea­ surements. For each sample run, 120 s worth of mass spectra were captured. The instrument’s settings of 2.20 mbar for the drift­tube pressure, 60°C for the drift temperature, and 550 V for the drift voltage pro­ duced an electric field strength to number density ratio (E/N) of 120 Td. Every sample was examined twice. Internal calibration of ToF spectra was per­ formed off­line after dead time correction in order to achieve high mass resolution (Cappellin et al., 2011). The PTR­ToF­MS’s better resolution offers a sum formula and a rough identification of each mass peak found. The TofDaq programme (Tofwerk AG, Thun, Switzerland) was used to collect, record, and analyze the data. Data were expressed in ppbv using a process outlined by Lindigner and Jordan (Lindinger and Jordan, 1998). Finally, all the VOC data were fil­ tered using a threshold of 0.50 ppbv and by eliminat­ ing any signals that may be attributed to the chem­ Fig. 1 ­ Chromatographic profiles of analyzed compounds (A) Chromatogram representing the analysis of hydrox­ ymethylfurfural (HMF) and furfural (F) content in honey samples. The retention times and identified compound names are indicated. (B) Chromatogram illustrating the sugar analysis in honey samples. Retention times and identified sugar names are provided. Taiti et al. ‐ Characterization of Italian honeys 333 also calculated. PLS­DA analysis was performed using PLS­Toolbox v. 8.0.2 (Eigenvector Research Inc., West Eaglerock Drive, Wenatchee, WA) for MATLAB R2015b (Mathworks Inc., Natick, MA, USA). In addition, to study the relationships between the different samples as a function of different physi­ co­chemical variables, a Factor Analysis (FA) was applied, considering as factors the content of the dif­ ferent analyzed sugars (glucose, fructose, sucrose, maltose), pH, electrical conductivity (mS/cm), and HMF (5­hydroxymethylfurfuraldehyde) level. This last parameter is essential to evaluate the compliance of honey with current legislation. The level of HMF is used as an indicator of the heating or high tempera­ ture storage of the honey. In fact, it is generally not present in fresh honey, while its content increases during conditioning and storage (Zappalà et al., 2005). Furthermore, it is inversely proportional to the fructose content and the fructose/glucose ratio (Kesić et al., 2014). Factor Analysis (FA) allows to visualize variables and samples simultaneously in a two or three­dimen­ sional space and to study the relationships between the observations (honey samples) and the variables (Greenacre, 1984; Escofier and Pagès, 1992). Computations were performed by XLSTAT Version 2014.5.03. 3. Results and Discussion pH The pH values of the examined honey samples fell within the acidic range, varying between 3.5 and 6, reported in Table 2 as the mean and the standard deviation (mean ± SD). Acacia honey exhibited the lowest average pH of 3.77 ± 0.13, closely followed by citrus honey with an average pH of 3.90 ± 0.51. istry of the water or to interfering ions, which are thought to be challenging to precisely quantify. Statistics were applied to the filtered data. Statistical analysis Multivariate partial least square­discriminant analysis (PLSDA) (supervised method) was applied to the spectra obtained from 84 honey samples pro­ duced by different genotypes, comprehensive of 38 protonated masses, for exploring the possibility of correctly classifying the botanical origin of the hon­ eys (acacia, chestnut, citrus fruits, linden, and wild­ flower, this last coming from a mix of species). As a pre­processing step, data were submitted to logarith­ mic transformation and auto­scaling. The whole data set was split into training and validation subset, opti­ mally chosen with the Euclidean distances based on the algorithm of Kennard and Stone (1969). The training set consisted of about 85% of the samples, used for selection of the optimal number of latent variables (LVs), model calibration and cross validation (internal validation). The test set, used to predict the class membership (external validation), included 15% of samples removed from the data set. The training set was used to build a model based on venetian blinds cross validation procedures, evaluated by the number of correct predictions and the root­mean­ square error of cross­validation (RMSECV), subse­ quently validated with the removed samples (exter­ nal validation set). External validation of the model was quantified by the root­mean­square error of pre­ diction (RMSEP). The optimal number of LVs was selected as those associated to the minimum error and misclassification rate of the calibration dataset. Confusion matrices were used to study the reliability of the models. The threshold to assign a sample to a class was chosen minimizing the number of false pos­ itives and false negatives (Bayes theorem). Variable Importance in Projection (VIP) scores (p = 0.01) were pH EC (mS/cm) Moisture content (%) HMF (mg/kg) F (mg/kg) Acacia 3.77 ± 0.13 0.27 ± 0.10 16.44 ± 0.74 0.95 ± 0.45 0.95 ± 0.54 Chestnut 5.19 ± 0.49 1.58 ± 0.33 16.60 ± 0.73 1.78 ± 0.48 2.21 ± 1.15 Citrus 3.90 ± 0.51 0.57 ± 0.54 17.20 ± 0.86 3.50 ± 1.64 2.74 ± 1.38 Linden 4.16 ± 0.06 0.92 ± 0.12 15.33 ± 1.15 2.40 ± 0.20 3.70 ± 1.53 Multifloral 4.12 ± 0.36 0.80 ± 0.42 16.16 ± 1.12 1.67 ± 0.56 1.53 ± 1.30 Aged 4.00 ± 0.19 0.70 ± 0.51 16.48 ± 1.85 6.82 ± 3.04 2.96 ± 1.17 Table 2 ­ Physico­chemical honey characteristics. The table reports the value of pH, electrical conductivity (EC), moisture content, hydroxymethylfurfural (HMF), and furfural (F) Data are reported as the mean ± standard deviation. 334 Adv. Hort. Sci., 2023 37(3): 329­341 Linden and multifloral honeys demonstrated slightly higher average pH values of 4.16 ± 0.06 and 4.12 ± 0.36, respectively. As expected, multifloral honeys displayed a considerable range of pH values (3.44 to 5.05), reflective of their inherent compositional diversity. In contrast, chestnut honey exhibited the highest pH value of 5.19 ± 0.49 among the tested varieties. Notably, the pH of aged honey, at 4.00 ± 0.19, aligned closely with the pH values of other hon­ eys of corresponding botanical origins, such as citrus and multifloral. The acidic nature of the honey sam­ ples has implications for their antimicrobial activity (Acquarone et al., 2007). The observed pH variations align with previous findings, with acacia and citrus honeys consistently displaying lower pH levels while chestnut the highest (Bertoncelj et al., 2011; Živkov­ Baloš et al., 2018). The near 4 pH level observed in other honey types is consistent with results reported in earlier studies (Truzzi et al., 2014). Electrical conductivity (EC) The EC of honey samples ranged between 0.1 and 2 mS/cm, depending by the botanical origin (Table 2). Acacia honeys were characterized by a considerably low EC (with an average value of 0.27 ± 0.10 mS/cm); in contrast, chestnut honeys had high EC values, with an average of 1.58 ± 0.33 mS/cm and a maximum value of 1.96 mS/cm. Citrus showed a mean of 0.57 ± 0.54 mS/cm, and linden of 0.92 ± 0.12 mS/cm. On the other hand, multifloral honeys exhibited a wide range of values, from 0.21 mS/cm to 1.86 mS/cm, the overall mean was 0.80 ± 0.42 mS/cm. Additionally, aged honey samples did not show significant differ­ ences when compared to honey of the same botani­ cal origin (0.69 ± 0.51 mS/cm). The EC values of most honey samples fell within the standard limit (Table 2), with the exception of two honeys. Notably, linden honey exhibited an average electrical conductivity (EC) that exceeded the established threshold. However, when considering a limit of 0.8 mS/cm, the EC values for linden honey (0.92 mS/cm) remained acceptable, thanks to an exemption stated in D.lgs. 179/04, which allows EC levels above 0.8 mS/cm. Nevertheless, EC is considered a reliable indicator of the botanical origin of honey. Chestnut honey is char­ acterised by a high EC value, followed by linden honey, which also displayed a relatively high value, as reported in other studies (Truzzi et al., 2014; Živkov­ Baloš et al., 2018). Excluding the outlier value of cit­ rus honey, the average EC of the samples was similar to findings in other studies (0.24 mS/cm) (Di Marco et al., 2017; Di Rosa et al., 2019). Conversely, multi­ floral honeys showed a wide range of EC values, ranging from 0.21 mS/cm to 1.86 mS/cm, reflecting the variation in floral sources visited by the bees. Moisture content The moisture content of all the honey types ranged from 13.4% to 19.6%, and all the samples were under the maximum limit of 20% (Table 2). Moreover, 94% of the samples meet the criteria for quality competitions (value ≤ 18%). The highest aver­ age value was 17.20 ± 0.86% of citrus honey, while the lower was 15.33 ± 1.15% of linden honey. Acacia, chestnut, multifloral, and aged honeys had similar values of 16.44 ± 0.74%, 16.60 ± 0.73%, 16.16 ± 1.12%, and 16.48 ± 1.85%, respectively. Honey mois­ ture content is an important factor and a parameter used to evaluate the product’s quality. Values that are too low can cause processing problems, while values that are too high could lead to the onset of fermentation processes, altering its quality, shelf life, taste, and composition (El Sohaimy et al., 2015). There was no discernible difference between the water content of various varieties of honey when the samples were compared, despite the significant vari­ ety and botanical origin of the samples. Indeed, there is a relationship between moisture content and honey maturation, production season, ventilation of the beehive, meteorological conditions and work processes (Kirs et al., 2011; Escuredo et al., 2014; De Sousa et al., 2016; Lazarević et al., 2017). Hydroxymethylfurfural (HMF) and furfural (F) The HPLC quantification of HMF (hydroxymethyl­ furfural) showed that all samples had HMF content within the standard thresholds. Acacia honey exhibit­ ed an average HMF content of 0.95±0.45 mg/kg, while chestnut honey showed a higher value of 1.78 ±0.48 mg/kg of HMF. Citrus honey recorded an even higher content, with 3.50±1.68 mg/kg of HMF with the higher value represented from the aged commer­ cial sample. Samples of linden honey displayed an average HMF content of 2.40±0.20 mg/kg, while mul­ tifloral honey showed a mean value of 1.67±0.56 mg/kg of HMF. However, the main difference was highlighted between fresh and aged honey. Indeed, samples of aged honey showed a considerable rise in HMF content, with an average of 6.82 ± 3.04 mg/kg. All honey harvested in 2022 had values from 0.5 to 3 mg/kg, while aged honey had significantly higher val­ ues from 4.6 to 12.12 mg/kg (Table 2). These results clearly indicate that ageing process can significantly influence HMF levels, which serve as Taiti et al. ‐ Characterization of Italian honeys 335 an important indicator of honey quality and fresh­ ness. Indeed, as reported in previous studies, these compounds are related to the heating practices and preservation conditions of honey and derive from the degradation of fructose (Aronne and De micco, 2010; Tedesco et al., 2022). In the same chromatographic run of HMF, furfural (F) data was also obtained. Acacia honey showed a mean of 0.95 ± 0.54 mg/kg, representing the honey with the lowest average F content, ranging from 0 to 1.89 mg/kg. Multifloral honey followed with a con­ tent of 1.53­1.30 mg/kg. Chestnut honey exhibited an average F content of 2.21 ± 1.15 mg/kg. Both citrus and aged honey displayed similar values of F: 2.74 ± 1.38 mg/kg and 2.96 ± 1.17 mg/kg, respectively. The highest mean value was found in linden honey (3.70 ± 1.53 mg/kg); however, the sample with the highest F content was multifloral honey with 6.74 mg/kg. Also, if neither restrictions nor indications are report­ ed for furfural in the legislation, it is related to stor­ age and of honey, since both F and HMF are usually produced by the Maillard reaction (Zhang et al., 2009). However, the average furfural content in the different honey types was in line with other studies (Gaspar and Lopes, 2009; Apriceno et al., 2018; Tedesco et al., 2022). Brix The degree Brix analysis, representing the total sugar content in honey, revealed that all honey sam­ ples exhibited values ranging from 78.8% to 85.5%. Among the varieties, citrus honey displayed the low­ est mean value (81.10±0.84%), while linden honey showcased the highest (82.93 ± 1.10%) (Table 3). Similarly, in line with previous studies, our Brix values were found to be comparable to those reported, reaffirming the absence of significant distinctions in sugar content among different honey botanical ori­ gins. However, similar to what was reported in earlier studies that also found similar brix values, the study of sugars using the refractometer did not reveal any appreciable differences between different types of honey (Oroian and Ropciuc, 2017; Geană et al., 2020). Sugar quantification Quantitative analysis of fructose, glucose, sucrose, and maltose was conducted using HPLC, with results expressed as percentages (g/g). Comprehensive data, including the sum of fructose and glucose, as well as individual sugar levels, are presented in Table 3. Acacia honey exhibited the highest fructose content at 49.08 ± 1.71%, while aged honey displayed the lowest fructose content (39.18 ± 3.50%). The fructose content across all samples ranged from 33.1% to 52.8%. Glucose content ranged from 22% to 40.2%, with chestnut honey demonstrating the lowest aver­ age value (27.24 ± 2.93%) and aged honey the high­ est (33.71 ± 4.76%). The range of maltose concentra­ tion was 0.9% to 4.8%, with chestnut honey having the highest level and linden honey the lowest (Table 3). Additionally, according to Council Directive 2001/110/CE of December 20, 2001, in unadulterat­ ed honeys, the sum of glucose and fructose should not fall below 60 g/100g for nectar honey, while sucrose must not exceed 5 g/100g. More specifically, 5g/100g for Acacia (Robinia pseudoacacia), Lucerne (Medicago sativa), Banksia (Banksia menziesii), Sulla (Hedysarum coronarium), Eucalyptus (Eucalyptus camaldulensis), not more than 10 g/100 g for Citrus (Citrus spp.) honey, and no more than 15 g/100 g for Lavender (Lavandula spp.) and Borage (Borago offici‐ nalis). In reference to these criteria, all 84 samples were unadulterated, in accordance with the legisla­ tion. The sum of glucose and fructose ranged from 63.80 to 84.90%, affirming the high quality of the honey samples, while sucrose was always lower than Brix Fructose Glucose Maltose G + F F/G Acacia 81.90 ± 0.72 49.08 ± 1.71 30.05 ± 2.08 3.12 ± 0.45 79.12 ± 2.68 1.64 ± 0.13 Chestnut 81.71 ± 0.77 45.34 ± 2.43 27.24 ± 2.93 3.36 ± 0.82 72.58 ± 4.76 1.68 ± 0.15 Citrus 81.10 ± 0.84 42.65 ± 2.57 32.79 ± 5.24 2.90 ± 1.02 75.44 ± 6.74 1.32 ± 0.17 Linden 82.93 ± 1.10 43.28 ± 2.02 28.39 ± 8.25 2.48 ± 0.80 71.66 ± 8.22 1.61 ± 0.42 Multifloral 82.19 ± 1.16 43.85 ± 3.71 30.26 ± 3.79 2.98 ± 0.87 74.11 ± 5.25 1.47 ± 0.23 Aged 82.24 ± 1.89 39.18 ± 3.50 33.71 ± 4.76 2.74 ± 0.62 72.89 ± 7.81 1.17 ± 0.10 Table 3 ­ The table reports the value (average and standard deviation) of degree Brix of honey (%) that represents the total sugar con­ tent and Fructose, Glucose, and Maltose expressed as percentages (g/g). In the table it is also reported the sum of Glucose and Fructose (G+F) and the Fructose Glucose ratio (F/G) Adv. Hort. Sci., 2023 37(3): 329­341 336 1.8%, detected in only 15 samples of the total, with a maximum value of 1.8% (7 acacia, 1 citrus and 7 mul­ tifloral). Moreover, the Fructose/Glucose (F/G) ratio was calculated. The F/G ratio determines whether honey may crystallise; therefore, a ratio higher than 1 suggests a fluid honey, whereas a ratio lower than 1 indicates honey crystallizing more quickly (Geană et al., 2020). The highest values were found in acacia, chestnut, and linden honeys (1.64, 1.67, and 1.60, respectively), indicating honey’s ability to remain liq­ uid for a longer amount of time. Citrus, multifloral flowers, and aged honeys, on the other hand, showed lower ratios (1.32, 1.47, and 1.17, respective­ ly). Additionally, no samples had a value lower than 1, however aged honey with an F/G ratio of 1.17 is most likely to have crystallized. In the current study, the fructose and glucose values found across differ­ ent honey samples are, on average, higher compared to those reported in other studies. However, the F/G ratio for acacia, citrus, and multifloral honeys remains consistent with literature values (Oddo and Piro, 2004; Geană et al., 2020). Factor analysis (FA) In figure 2, the FA biplot simultaneously repre­ sents the relationship between the different sugars analyzed (glucose, fructose, sucrose, maltose), the level of HMF, the pH and the electrical conductivity, highlighting the relative distances among the 84 honey samples. The first axis explains 37.18% of the total variability in the spectral data, the second axis 18.01%. From the FA graph, some groups of samples emerge which seem to be related to a compound or to a specific physico­chemical characteristic of the honey. In particular, the wildflowers are concentrat­ ed on the HMF vector, and, in a diametrically oppo­ site position, the samples of acacia honey are grouped very close along the fructose axis. This is in accordance with the fact that HMF is formed through the degradation of fructose, thus establishing a nega­ tive correlation between these two parameters. Citrus and Linden honeys are situated in the lower region of the graph, indicating slightly higher glucose values for citrus honey and higher HMF values for both honeys. The pH values are also significantly higher in chestnut samples, although maltose seems to somehow influence its distribution. PTR‐ToF‐MS results Data on the emissions of volatile organic com­ pounds (VOCs) from the five honey groups ­ Multifloral, Acacia, Chestnut, Citrus, and Linden­ are presented in Table 4. All signals have been separated according to their respective molecular weights and are expressed as the mean concentration in parts per billion by volume (ppbV). The Table shown a subset of 33 compounds obtained upon filtering the data (were eliminated all signals with an average concen­ tration below to 1 ppbV). From our study on 84 dif­ ferent honey samples a total of 37 different com­ pounds with an average value higher than 1ppbv were found. Among these, the peaks with the higher emission were detected at 33.034, 45.033, 47.010, 59.049, which corresponding to the following com­ pounds methanol, acetaldehyde, formic acid and ace­ tone. Similar results were obtained from other stud­ ies on honey samples from different botanical origins (Kuś and van Ruth, 2015; Schuhfried et al., 2016). The average of total emission recorded for each honey botanical origin varies from a minimum value of 265.2 ppbv for Acacia and 336.1 for Linden honey to a maximum value of 1971.8 for citrus honey (Table 4). Acacia and Linden honey showed a rather similar volatile profile characterized by both a lower level of emission and a lower number of signals (30 and 28 respectively) compared to the other botanical ori­ gins. Citrus honey emerged both for a higher emis­ sion of methanol, acetaldehyde and acetone com­ pounds as well for a higher emission of terpene com­ pounds (mz 111.101, 121.101, 135.116, 137.132) compared to the other botanical origins. Chestnut samples are characterized by large signals of com­ pounds detected at m/z: 69.033 (C4H5O+, Tentatively Identified as Furan), 83.086 (C6H11+, TI as C6 compounds) and 105.069 (C5H13S+, TI pen­ Fig. 2 ­ Biplot from Factor analysis. Relationships between 84 honey samples and different physicochemical parame­ ters. AC = acacia, CI = citrus fruits, CH = chestnut, WF = wildflower, LI = linden. Taiti et al. ‐ Characterization of Italian honeys 337 Table 4 ­ Number of signals detected, chemical formula, tentative identification, VIP score and average amount (ppbV) of each com­ pound detected from different honey samples N° of compounds mz Chemical Compound Tentative Identification* Multifloral (n=49) Acacia (n=16) Chestnut (n=11) Citrus (n=5) Linden (n=3) 1 27022 C2H3+ Acetylene 33.49 10.80 30.91 99.50 11.85 2 33033 CH5O+ Methanol 189.78 61.30 61.15 243.54 49.04 3 41038 C3H5+ Alkyl fragment 22.58 5.39 11.16 50.39 19.03 4 43018 C2H3O+ aceton or acetate fragments 44.33 14.86 47.49 87.52 23.39 5 45033 C2H5O+ Acetaldehyde** 369.53 88.24 506.01 884.92 106.64 6 47010 CH3O2+ Formic acid/formats** 25.21 7.05 41.97 180.30 7.90 7 49011 CH5S+ S compound (Methanethiol) 3.23 2.10 2.93 8.96 2.00 8 53038 C4H5+ Cyclobutadiene 2.48 2.03 2.21 2.79 2.00 9 55054 C4H7+ Alkyl fragment 6.43 4.09 5.93 18.59 3.68 10 57069 C4H9+ Alcohol fragments 13.25 5.02 3.02 6.33 2.59 11 59049 C3H7O+ Acetone** 119.70 42.56 210.57 284.23 100.21 12 61028 C2H5O2+ Acetic acid 12.25 7.25 11.14 15.98 5.94 13 63033 C2H7S+ S compound (Dimethyl sulfide) 13.07 4.07 17.29 25.27 2.15 14 65.00 C5H5+ S compound 2.59 Tr 2.35 3.63 Tr 15 67050 C5H7+ 3­Penten­1­yne/Terpene fragment Tr Tr Tr 2.65 Tr 16 69033 C4H5O+ Furan 5.33 2.93 16.02 5.84 Tr 17 71086 C5H11+ Alcohol compounds 2.24 2.77 2.24 1.10 1.10 18 73054 C4H9O+ Butan­2­one 13.60 5.54 10.70 20.21 3.48 19 75044 C3H7O2+ Isobutanol 2.67 2.71 2.58 2.64 2.33 20 77038 C6H5+ Alkyl fragment 2.11 2.03 2.06 2.14 1.01 21 79049 C6H7+ Benzene/terpene fragment 2.88 2.72 2.52 3.34 0.00 22 83086 C6H11+ C6 compounds 2.35 2.02 6.92 2.87 1.10 23 85059 C5H9O+ (E)­2­Pentenal 3.14 2.17 2.14 3.21 0.00 24 87044 C4H7O2+ 2,3­Butanedione 7.59 3.65 10.10 15.91 3.81 25 93069 C7H9+ Terpene fragments Tr Tr Tr 3.23 Tr 26 95011 C2H7O2S+ dimethyl sulfone 7.34 4.53 3.52 6.44 3.82 27 97033 C5H5O2+ Furfural 6.09 3.26 4.30 8.49 2.63 28 99044 C5H7O2+ Furfuryl alcohol Tr Tr Tr Tr Tr 29 101060 C5H9O2+ Dihydro­methyl­furanone Tr Tr 0.00 0.00 0.00 30 103080 C5H11O2+ 2­/3­methylbutyric acid Tr 0.00 0.00 2.71 0.00 31 105069 C5H13S+ Pentanethiol** Tr 0.00 4.26 1.89 Tr 32 107086 C8H1 1+ 1,3­Dimethylbenzene/Terpenes fragments Tr Tr 1.87 1.58 0.00 33 109070 C7H9O+ Benzyl alcohol Tr 0.00 Tr 1.12 1.41 34 111101 C8H15+ Terpenes fragments** Tr 0.00 Tr 1.88 0.00 35 121101 C9H1 3+ Terpenes fragments Tr 0.00 1.12 2.54 0.00 36 135116 C10H15+ Terpenes** Tr 0.00 Tr 2.10 0.00 37 137132 C10H17+ Terpenes Tr 0.00 0.00 2.08 0.00 Total emission average 889.30 265.28 996.51 1971.86 332.26 Total signals detected 37 30 34 36 28 * Each value has been tentatively assigned to compounds, based on PTR­honey literature data (Kuś and van Ruth, 2015; Schuhfried et al., 2016; Ballabio et al., 2018). ** compounds with the highest VIP score. Tr means trace and these compounds have been identified in at least 1 sample per group but with an overall average value below 1 ppbV. Adv. Hort. Sci., 2023 37(3): 329­341 338 tanethiol) in agreement with Ballabbio et al. (2018). As can be seen from the data shown in the table 4, the average volatile profile of multifloral honey showed an average emission for many signals and trace compounds in large numbers (identified only in some samples). No significant differences were observed between the volatile organic compound (VOC) emissions of commercial honey and those produced by beekeep­ ers. PLS‐DA analysis With the aim to get an overview of the VOC data collected, a PLS­DA analysis was applied on the whole dataset obtained from 37 different VOCs data collect­ ed from 84 samples. It emerges that the honey sam­ ples distance themselves from each other according to their botanical origin. Multifloral honey seem to show a variable trend probably linked to its botanical origin. To provide a more detailed characterization of the VOCs emitted by different honey samples, VIP scores higher than 1 and their possible identification on the basis of literature data were reported in Table 4 (marked by two asterisks). The volatile compounds with higher VIP value could be good candidates for the honey species identification. In particular, the chemical species with the higher significance were detected at mz 45.033(TI Acetaldehyde), 47.01 (TI Formic acid/formates), 59.049 (TI Acetone), 105.069 (TI Pentanethiol), 111.10 (TI Terpenes fragments), 135.116 (TI Terpenes). PLS­DA approach was applied to find VOCs able to discriminate among species. By applying the model developed by the PLS­DA on honey samples of differ­ ent botanical origin, a correct distinction of the taxo­ nomic category of two/five different groups was achieved. Indeed, the multifloral honey could be obtained by honeybees from the nectar of different flowers. Score plot from the PLS­DA model is shown in figure 3. The global quality of the model, evaluated by its performances indicators (Table 5), resulted robust enough to discriminate the botanical origin of the citrus and chestnut samples compared to the others in the calibration/validation data set, and in the independent test set. Indeed, the PLS­DA three­ component model successfully classified 100% of Fig. 3 ­ Score plot (LV1, LV2) of the PLS­DA model. Samples of different botanical origin are highlighted. Red = acacia; green = citrus fruits; blue = chestnut; light blue = wild­ flower; lilac = linden. Statistics Y­BLOCKS Class 1 ­ acacia Class 2 – citrus fruits Class 3 ­ chestnut Class 4 ­ wildflower Class 5 ­ linden Sensitivity (SE) (Cal) 1.000 1.000 1.000 0.721 1.000 Specificity (SP) (Cal) 0.800 1.000 1.000 0.724 0.643 Sensitivity (SE) (CV) 0.833 1.000 1.000 0.744 1.000 Specificity (SP) (CV) 0.800 0.970 1.000 0.552 0.643 Sensitivity (SE) (P) 1.000 0.000 0.000 0.000 1.000 Specificity (SP) (P) 0.875 1.000 0.833 0.182 0.583 Class. error (Cal) 0.100 0.000 0.000 0.277 0.178 Class. error (CV) 0.183 0.015 0.000 0.352 0.178 Class. error (Pred) 0.062 0.500 0.583 0.909 0.208 RMSEC 0.319 0.104 0.169 0.414 0.161 RMSECV 0.328 0.159 0.199 0.464 0.166 RMSEP 0.387 0.296 0.719 0.674 0.044 Table 5 ­ PLS­DA statistics for the honey samples for the five botanical origins: 1 = acacia, 2 = citrus fruits, 3 = chestnut, 4 = wildflower, 5 = linden. Sensitivity (SE), Specificity (SP), Class. Error, RMSEC, RMSECV, RMSEP, for Calibration (Cal), Cross Validation (CV), and Prediction (Pred), respectively Taiti et al. ‐ Characterization of Italian honeys 339 honey samples from classes 2­chestnut and 3­citrus into their respective taxonomic categories during fit­ ting, cross­validation (internal validation), and predic­ tion (external validation), while the acacia and linden honey samples are confused with the multifloral samples. 4. Conclusions In this study, a comprehensive analysis of various physico­chemical properties and volatile organic compounds (VOCs) present in Italian honeys was con­ ducted. It was revealed that the quality of the honey sold is excellent, as legal limits were adhered to for all samples (except for few EC values). The electrical conductivity (EC) values demonstrated significant variability, with chestnut and linden honeys standing out due to their high and relatively high EC values, respectively. The observed pH variations among dif­ ferent honey types were consistent with their botani­ cal origins. Furthermore, parameters such as sucrose content and the fructose­to­glucose ratio, indicative of potential adulterations, remained within legal lim­ its for both commercial and beekeeper honey. Moreover, within this context, the honey varieties were discerned based on their distinctive characteris­ tics. Aged honey, as expected, was characterized by a high HMF level; however, it remained within legal thresholds. Acacia honey, characterized by its high fructose content, exhibited low pH and EC values. Interestingly, in VOC analysis, it displayed similari­ ties with linden honey due to their shared low emis­ sions of volatile compounds. Chestnut honey, which had high pH and EC values, was easy to differentiate from other types of honey using both conventional metrics and PTR­ToF­MS­based VOC analyses. Its distinctive profile made it simple to classify. Citrus honey displayed physicochemical characteristics similar to linden and multifloral honeys, but its dis­ tinctive VOC emissions allowed for a more accurate identification. Conversely, linden and multifloral honeys shared close resemblances in chemical and physical analyses, and the significant variability in the multifloral variety’s composition due to its diverse floral sources hindered differentiation through VOCs. Factor Analysis provided insights into the rela­ tionships between different sugars, HMF, pH, and electrical conductivity, highlighting distinct group­ ings of honey samples. Finally, VOCs analysis revealed a diverse range of compounds, with noticeable variations attributed to the botanical ori­ gin of the honey. Partial Least Squares­Discriminant Analysis (PLS­DA) facilitated discrimination among different honey types based on their VOC profiles highlighting the potential for this method in distin­ guishing honey types based on VOC profiles. The analysis of VOCs has the advantage of being a faster alternative to pollen analysis, providing an efficient means of differentiating between honey samples with varying botanical sources. For this reason, PTR­ ToF­MS­based VOC analysis serves as a valuable tool to complement or even replace melissopalyno­ logical analysis, as demonstrated by the effective combination of VOC analysis and PLS­DA for distin­ guishing honeys of different botanical origins. In conclusion, the collective application of physi­ co­chemical and VOC analyses yielded a compre­ hensive means of effectively characterizing honey varieties. This integrated approach underscores the robustness of employing multiple techniques for a thorough understanding of honey attributes. Acknowledgements This research is part of the BEEWIN project: BEEkeepers Weather indexed INsurance project funded by MIPAAF (Ministero delle Politiche Agricole Alimentari e Forestali). References ACQUARONE C., BUERA P., ELIZALDE B., 2007 ­ Pattern of pH and electrical conductivity upon honey dilution as a complementary tool for discriminating geographical origin of honeys’. ­ Food Chem., 101(2): 695­703. APRICENO A., GIRELLI A.M., SCUTO F.R., TAROLA A.M., 2018 ­ Determination of furanic compounds and acidity for Italian honey quality. ­ Flavour Fragrance J., 33(6): 411­419. ARONNE G., DE MICCO V., 2010 ­ Traditional melissopaly‐ nology integrated by multivariate analysis and sam‐ pling methods to improve botanical and geographical characterisation of honeys. ­ Plant Biosystems, 144(4): 833­840. BALLABIO D., ROBOTTI E., GRISONI F., QUASSO F., BOBBA M., VERCELLI S., .GOSETTI F., CALABRESE G., SANGIOR­ GI E., ORLANDI M., MARENGO E., 2018 ­ Chemical pro‐ filing and multivariate data fusion methods for the Adv. Hort. Sci., 2023 37(3): 329­341 340 identification of the botanical origin of honey. ­ Food Chem., 266: 79­89. BERTONCELJ J., GOLOB T., KROPF U., KOROŠEC M., 2011 ­ Characterisation of Slovenian honeys on the basis of sensory and physicochemical analysis with a chemo‐ metric approach. ­ Inter. J. Food Sci. Techn., 46(8): 1661­1671. BLAKE R.S., MONKS P.S., ELLIS A.M., 2009 ­ Proton‐Transfer Reaction Mass Spectrometry. ­ Chem. Reviews, 109(3): 861­896. CAPPELLIN L., BIASIOLI F., GRANITTO P.M., SCHUHFRIED E., SOUKOULIS C., COSTA F., MÄRK T.D., GASPERI F., 2011 ­ On data analysis in PTR‐TOF‐MS: From raw spectra to data mining. ­ Sensors Actuators B, Chem., 155(1): 183­ 190. CASTIGLIONI S., STEFANO M., ASTOLFI P., CARLONI P., 2017 ­ Chemometric approach to the analysis of antiox‐ idant properties and colour of typical Italian monofloral honeys. ­ Inter. J. Food Sci. Technol., 52(5): 1138‐1146. COUNCIL EUROPEAN UNION, 2001 ‐ Directive 2001/110/EC, (2001) relating to honey, ­ Official J. European Communities, EN, L 10/47­52 http://www. ihc­platform.net/honeydirective2001.pdf. DE SOUSA J.M.B., DE SOUZA E.L., MARQUES G., DE TOLE­ DO BENASSI M., GULLÓN B., PINTADO M.M., MAGNANI M., 2016 ­ Sugar profile, physicochemical and sensory aspects of monofloral honeys produced by different stingless bee species in Brazilian semi‐arid region. ­ LWT ­ Food Sci. Techn., 65: 645­651. DI MARCO G., MANFREDINI A., LEONARDI D., CANUTI L., IMPEI S., GISMONDI A., CANINI A., 2017 ­ Geographical, botanical and chemical profile of monofloral Italian honeys as food quality guarantee and territory brand. ­ Plant Biosystems, 151(3): 450­463. DI ROSA A.R., LEONE F., CHELI F., CHIOFALO V., 2019 ­ Novel approach for the characterisation of sicilian hon‐ eys based on the correlation of physico‐chemical para‐ meters and artificial senses. ­ Italian J. Animal Sci., 18(1): 389­397. EL SOHAIMY S.A., MASRY S.H.D., SHEHATA M.G., 2015 ­ Physicochemical characteristics of honey from different origins. ­ Annals Agric. Sci., 60(2): 279­287. ESCOFIER B., PAGÈS J., 1992 ­ Análisis factoriales simples y múltiples: Obejtivos, métodos e interpretación. ­ Universidad del país Vasco, Bilbao, Spain, pp. 286. GREENACRE M.J., 1984 ­ Theory and applications of corre‐ spondence analysis. ­ Academic Press Inc., London, UK, pp. 364. ESCUREDO O., DOBRE I., FERNÁNDEZ­GONZÁLEZ M., SEIJO M.C., 2014 ­ Contribution of botanical origin and sugar composition of honeys on the crystallization phenome‐ non. ­ Food Chem., 149: 84­90. FALLICO B., ZAPPALA M., ARENA E., VERZERA A., 2004 ­ Effects of conditioning on HMF content in unifloral hon‐ eys. ­ Food Chem., 85(2): 305­313. GASPAR E.M.S.M., LOPES J.F., 2009 ­ Simple gas chromato‐ graphic method for furfural analysis . ­ J. Chromatography A, 1216(14): 2762­2767. GAZZETTA UFFICIALE, 2003 ­ Approvazione dei metodi uffi‐ ciali di analisi da applicarsi per la valutazione delle caratteristiche di composizione del miele. ­ https://www.gazzettaufficiale.it/eli/id/2003/08/11/03A 09206/sg. GEANĂ E.I., CIUCURE C.T., COSTINEL D., IONETE R.E., 2020 ­ Evaluation of honey in terms of quality and authentici‐ ty based on the general physicochemical pattern, major sugar composition and δ13C signature. ­ Food Control, 109: 106919. ISMEA, 2023 ­ Api e Miele ‐ Analisi e studio filiera miele. ­ Ismea Mercati, https://www.ismeamercati.it/api­miele. KENNARD R., STONE L., 1969 ­ Computer aided design of experiments on JSTOR . ­ https://www.jstor.org/stable/1266770. KESIĆ A., CRNKIĆ A., HODŽIĆ Z., IBRIŠIMOVIĆ N., ŠESTAN A., 2014 ­ Effects of botanical origin and ageing on HMF content in bee honey. ­ J. Sci. Res. Reports, 3(8). 1057­1066. KIRS E., PALL R., MARTVERK K., LAOS K., 2011 ­ Physicochemical and melissopalynological characteriza‐ tion of Estonian summer honeys. ­ Procedia Food Sci., 1: 616­624. KUŚ P.M., VAN RUTH S., 2015 ­ Discrimination of Polish unifloral honeys using overall PTR‐MS and HPLC finger‐ prints combined with chemometrics. ­ LWT ­ Food Sci. Techn., 62(1): 69­75. LAZAREVIĆ K.B., JOVETIĆ M.S., TEŠIĆ Ž.L., 2017 ­ Physicochemical parameters as a tool for the assess‐ ment of origin of honey. ­ J. AOAC Inter., 100(4): 840­ 851. LINDINGER W., JORDAN A., 1998 ­ Proton‐transfer‐reaction mass spectrometry (PTR–MS): on‐line monitoring of volatile organic compounds at pptv levels’. ­ Chem. Soci. Rev., 27(5): 347­375. MUREȘAN C.I., CORNEA­CIPCIGAN M., SUHAROSCHI R., ERLER S., MĂRGĂOAN R., 2022 ­ Honey botanical origin and honey‐specific protein pattern: Characterization of some European honeys. ­ LWT ­ Food Sci. Techn., 154: 112883. NIKHAT S., FAZIL M., 2022 ­ History, phytochemistry, exper‐ imental pharmacology and clinical uses of honey: A comprehensive review with special reference to Unani medicine. ­ J. Ethnopharmacol., 282: 114614. ODDO L.P., PIRO R., 2004 ­ Main European unifloral hon‐ eys: descriptive sheets. ­ Apidologie, 35(1): S38­S81. OROIAN M., ROPCIUC S., 2017 ­ Honey authentication based on physicochemical parameters and phenolic compounds. ­ Computers Electronics Agric., 138: 148­ 156. SCHUHFRIED E., DEL PULGAR J.S., BOBBA M., PIRO R., CAP­ PELLIN L., MÄRK T.D., BIASIOLI F., 2016 ­ Classification of 7 monofloral honey varieties by PTR‐ToF‐MS direct headspace analysis and chemometrics. ­ Talanta, 147: https://biblioteca.ucatolica.edu.co/cgi-bin/koha/opac-search.pl?q=pb:Universidad%20del%20%20pa%C3%ADs%20Vasco,%20 Taiti et al. ‐ Characterization of Italian honeys 341 213­219. TEDESCO R., SCALABRIN E., MALAGNINI V., STROJNIK L., OGRINC N., CAPODAGLIO G., 2022 ­ Characterization of botanical origin of Italian honey by carbohydrate com‐ position and volatile organic compounds (VOCs). ­ Foods, 11(16): 2441. TRUZZI C., ANNIBALDI A., ILLUMINATI S., FINALE C., ROS­ SETTI M., SCARPONI G., 2012 ­ Determination of very low levels of 5‐(hydroxymethyl)‐2‐furaldehyde (HMF) in natural honey: Comparison between the HPLC tech‐ nique and the spectrophotometric white method. ­ J. Food Sci., 77(7): C784­C790. TRUZZI C., ILLUMINATI S., ANNIBALDI A., FINALE C., ROS­ SETTI M., SCARPONI G., 2014 ­ Physicochemical proper‐ ties of honey from Marche, Central Italy: classification of unifloral and multifloral honeys by multivariate analysis . ­ Nat. Product Commun., 9(11): 1934578X1400901117. VÎJAN L.E., MAZILU I.C., ENACHE C., ENACHE S., TOPALĂ C.M., 2023 ­ Botanical origin influence on some honey physicochemical characteristics and antioxidant proper‐ ties. ­ Foods, 12(11): 2134. WARUI M.W., HANSTED L., GIKUNGU M., MBURU J., KIRONCHI G., BOSSELMANN A.S., 2019 ­ Characterization of Kenyan honeys based on their physicochemical properties, botanical and geographi‐ cal origin. ­ Inter. J. Food Sci., 2019: 932509. ZAPPALA M., FALLICO B., ARENA E., VERZERA A., 2005 ­ Methods for the determination of HMF in honey: a comparison. ­ Food Control, 16(3): 273­277. ZHANG Q., AMES J.M., SMITH R.D., BAYNES J.W., METZ T.O., 2009 ­ A perspective on the Maillard reaction and the analysis of protein glycation by mass spectrometry: probing the pathogenesis of chronic disease. ­ J. Proteome Res., 8(2): 754­769. ZHOU Q., WINTERSTEEN C.L., CADWALLADER K.R., 2002 ­ Identification and quantification of aroma‐active com‐ ponents that contribute to the distinct malty flavor of buckwheat honey. ­ J. Agric. Food Chem., 50(7): 2016­ 2021. ŽIVKOV­BALOŠ M., POPOV N., VIDAKOVIĆ S., LJUBOJEVIĆ­ PELIĆ D., PELIĆ M., MIHALJEV Ž., JAKŠIĆ S., 2018 ­ Electrical conductivity and acidity of honey. ­ Archiv. Veter. Med., 11(1): 91­101.