A new model to identify botanical origin of Polish honeys based on the physicochemical parameters and chemometric analysis

Accepted Manuscript A new model to identify botanical origin of polish honeys based on the physicochemical parameters and chemometric analysis Stanisław Popek, Michał Halagarda, Karolina Kursa PII:

S0023-6438(16)30780-0

DOI:

10.1016/j.lwt.2016.12.003

Reference:

YFSTL 5899

To appear in:

LWT - Food Science and Technology

Received Date: 17 May 2016 Revised Date:

1 December 2016

Accepted Date: 2 December 2016

Please cite this article as: Popek, S., Halagarda, M., Kursa, K., A new model to identify botanical origin of polish honeys based on the physicochemical parameters and chemometric analysis, LWT - Food Science and Technology (2017), doi: 10.1016/j.lwt.2016.12.003. This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

ACCEPTED MANUSCRIPT 1

A new model to identify botanical origin of Polish honeys based on the physicochemical

2

parameters and chemometric analysis

4

Stanisław Popeka

5

Michał Halagardab

6

Karolina Kursac

9 10 11

Department of Food Commodity Science, Cracow University of Economics, 30-033 Cracow,

Sienkiewicza 5, Poland, e-mail: [email protected] b

SC

8

a

Department of Food Commodity Science, Cracow University of Economics, 30-033 Cracow,

Sienkiewicza 5, Poland, e-mail: [email protected] c

Department of Food Commodity Science, Cracow University of Economics, 30-033 Cracow,

Sienkiewicza 5, Poland

13

Abstract

TE D

12

14

M AN U

7

RI PT

3

Confirmation of the authenticity or adulteration detection is a difficult, laborious and

16

costly process. The aim of this study was to construct a honey botanical origin classification

17

model on the basis of its characteristic physicochemical features. The experimental material

18

comprised of 72 samples of varietal honeys. The botanical origin and the purity of honey

19

samples were verified using a savoriness profiling method, and a pollen analysis. The

20

following parameters of chosen honeys were determined: water, total ash, reducing sugar,

21

total sugar and sucrose content, pH, total acidity solutions, specific electrical conductivity,

22

dynamic viscosity, diastatic number, 5-HMF and proline content. The classification model

23

was constructed with the use of all variables and with an employment of C&RT data mining

24

method. A v-fold cross-validation proved that the model does reproduce the structure of

AC C

EP

15

1

ACCEPTED MANUSCRIPT dataset very well and in this particular case, incorrectly classifying only in one case heather

26

honey as a multifloral one.

27

Keywords: honey; identification of honey; DATA-MINING C&RT; physicochemical

28

parameters of honey

RI PT

25

29 30

1. Introduction

Authenticity verification of honeys is linked with their varietal identification

32

(Bogdanov & Gallmann, 2008; Hastie, Tibshirani & Friedman, 2009). For many years, the

33

varietal identification of honey has been a research subject in many scientific centers

34

(Bogdanov & Gallmann, 2008). Pollen analysis has been the most often applied method when

35

identifying varietal honey. It is a traditional method used to confirm the biological origin of

36

honey elaborated and proposed by the Interantional Commission for Bee Botany (ICBB) in

37

1970, and later revised and updated in 1978 (Louveaux, Maurizio, & Vorwohl, 1978).

38

Nevertheless, this method is highly time-consuming and its accuracy depends strongly on the

39

skills of professional experts (Karabagias, Badeka, Kontakos, Karabournioti & Kontominas,

40

2014). Over recent years, this method has also been applied together with advanced statistical

41

methods. Honey bees collect floral pollen from various plants, thus, pure mono-pollen honeys

42

are very rarely encountered. Since the content of various pollens in honey is very large,

43

currently this method is used collaterally with the sensory and physicochemical analyses to

44

increase accuracy of the test (Bogdanov & Gallmann, 2008). Nevertheless, Stephens et al.

45

(2010) reported that melissopalynological analysis was of no practical use in differentiating

46

between manuka and kanuka honeys from New Zealand. However, identification of DNA

47

markers present in pollen to specifically confirm the botanical origin of honey is a novel and

48

promising method (Soares, Amaral, Oliveira & Mafra, 2015).

AC C

EP

TE D

M AN U

SC

31

2

ACCEPTED MANUSCRIPT Sensory analysis on its own is also used in the varietal identification of honey

50

(Kerlvliet, 1992). This method, however, is considered as subjective. Therefore, many

51

researchers endeavored to use the analysis of physicochemical parameters of honey in the

52

varietal and geographical identification of honey, instead or additionally (Persano Oddo et al.,

53

2004; Ruoff et al., 2007).

RI PT

49

Amidst all the physicochemical parameters of honey quality, specific electrical

55

conductivity appears to be the most effective when identifying varietal honey. This parameter

56

is mainly used to distinguish some varietal nectar honeys from nectar blossom and honeydew

57

honeys (Popek, 1998). Researches concerning honey identification on the basis of dyes

58

contents in honey, mainly flavonoids (Meda, Lamiec, Romito, Millogo & Nacoulma, 2005),

59

or of color parameters’ measurements in L* a* b* and X Y Z systems (Castro, Escamilla &

60

Reig, 1992) did not bring anticipated results. Multi-element analysis by inductively coupled

61

plasma optical emission spectroscopy (ICP-OES) and by inductively coupled plasma mass

62

spectrometry (ICP-MS) showed negative results as well (Di Bella et al., 2015).

TE D

M AN U

SC

54

Another approach concerns chemometric analysis of physicochemical parameters such

64

as: contents of: saccharides, nitrogen, sucrose, 5-hydroxymethylfurfural, ash, water, aromatic

65

acids and amino acids or glucose to fructose concentration ratio, pH level, total acidity,

66

specific rotation (Ojeda de Rodrıǵ uez, Sulbarán de Ferrer, Ferre & Rodrıǵ uez, 2004; Serrano,

67

Villareho, Espejo & Jodral, 2004). The combination of the above mentioned parameters

68

makes it possible to recognize some of the monofloral honeys (Terrab, González, Díez &

69

Heredia, 2003). Yet, the results of those analyses are not satisfactory. It is, though, not

70

possible on their basis to classify all honeys according to their individual types and varieties.

AC C

EP

63

71

Some of the volatile fractions determination methods facilitate the discrimination of

72

honey of different botanical origin. Escriche, Kadar, Juan-Borras and Domenech (2011)

73

proved the usefulness of flavonoids, phenolic compounds and headspace volatile profile

3

ACCEPTED MANUSCRIPT together with statistical data evaluation techniques (PCA and PLS2) for verification of the

75

botanical origin of lemon and orange honeys. Aliferis, Tarantilis, Harizanis and Alissandakis

76

(2010) proved that HS-SPME–GC/MS fingerprinting of honey volatiles combined with state-

77

of-the-art chemometrics (OPLS™-DA) provides a potential honey origin discrimination tool.

78

Additionally, in their research some biomarkers were detected. The existence of certain

79

marker compounds useful for selected honeys’ origin verification was also confirmed with

80

GC–MS aroma compounds analysis performed by Castro-Vázquez, Leon-Ruiz, Alañon,

81

Pérez-Coello and González-Porto (2014). According to Špánik, Pažitná, Šiška and Szolcsányi

82

(2014), the GC-MS evaluation of differences in distribution of enantiomers of chiral volatile

83

organic compounds holds a potential for distinguishing botanical origin of honey. The results

84

of the Gašić et al. (2015) study showed that the analysis of the phenolic characteristics of

85

honey achieved using UHPLC DAD–MS/MS as well as sugar and sugar alcohols determined

86

by HPAEC/PAD and mineral content specified with the use of ICP-OES has a significant

87

potential for the characterization of honey typical for a certain area. Zhao et al. (2016) proved

88

that by establishing chromatographic fingerprints with the use of HPLC–ECD it is possible to

89

identify three types of monofloral honey (Chinese jujube, longan and chaste). Moreover, a

90

chemometric analysis of selected volatile compounds and physicochemical parameters

91

(Karabagias, Badeka, Kontakos, Karabournioti & Kontominas, 2014) or selected phenolic

92

compounds and conventional physicochemical parameters (Karabagias et al., 2014) proved

93

useful in identifying the botanical origin of Greek honey.

SC

M AN U

TE D

EP

AC C

94

RI PT

74

Research by Wei and Wang (2014) showed that potentiometric and voltammetric

95

electronic tongues together with discriminant function analysis (DFA) are useful for

96

discrimination of monofloral honeys. Scandurra, Tripodi & Verzera (2013) used electrical

97

impedance spectroscopy to determine the botanical origin of monofloral honeys. They

98

showed that few parameters, such as parallel resistance and impedance of the circuit, can be

4

ACCEPTED MANUSCRIPT 99

used as indicators of the floral origin.

The application of near-infrared and mid-infrared

spectroscopy (Chen et al., 2012; Escuredo, González-Martín, Rodríguez-Flores, & Seijo,

101

2015), that involves analysis of spectra, enables distinguish honeydew honey from nectar

102

honey. The research carried out by Tewari and Irudayaraj (2004) proved that also FT-MIR

103

spectroscopy could be successfully applied when assessing the authenticity of food products

104

including honey.

RI PT

100

105

A further developed method is an isotope ratio mass spectroscopy (IRMS), which. is

107

based on the knowledge of the ratio of isotopes that are characteristic for individual plant

108

species. (Kelly, 2003).Whereas, according to Spiteri et al. (2015) also

109

measurements together with Independent Component Analysis (ICA) can be used to identify

110

specific markers that are typical of botanical origin of selected honeys.

M AN U

SC

106

1

H-NMR

Nevertheless, repeatedly the results of above mentioned, often complex analyses, that

112

in many cases require expensive equipment, do not permit the authenticity of honey to be

113

unfailingly confirmed or contested. Thus, other methods are suggested. They involve

114

measurements of a couple or a dozen physicochemical characteristics of honey and

115

chemometric analyses of the parameters measured (e.g.: variation analysis, canonical analysis,

116

analysis of key components, multidimensional analysis, taxonomic analysis, and discriminant

117

analysis) in order to reduce the complexity and to provide a better interpretation of data sets

118

and consequently to choose a few characteristics of honey that could be then considered for an

119

optimal varietal or geographic distinguishing feature (Terrab, González, Díez & Heredia,

120

2003; Yücel & Sultanoglu, 2013).

AC C

EP

TE D

111

121

The literature confirms the fact that although so many diverse honey identifying

122

methods are applied, it is still necessary to improve them or to develop a more effective

123

method to better and more successfully identify the type and the variety of honey. Therefore

5

ACCEPTED MANUSCRIPT 124

the aim of this study was to construct a honey classification model on the basis of its

125

characteristic physicochemical features to enable confirmation of the botanical origin of

126

honey.

127

2. Material and methods

RI PT

128 129

The experimental material comprised of 72 samples of varietal honeys (nectar [from

131

rape, acacia, heather, linden, buckwheat, and multifloral nectar from various plants]

132

honeydew, and nectar-honeydew). The honey samples analyzed were produced in apiaries

133

located throughout various regions of Poland: Lesser Poland (Małopolska) (19 samples),

134

Warmia and Masuria (13 samples), Silesia (13 samples), Pomerania (14 samples) and

135

Podlasie (13 samples). Each sample was acquired from a different apiary.

M AN U

SC

130

The botanical origin and the purity of honey samples were verified using a savouriness

137

profiling method developed by Cairnocros and Sjőstrőm and modified by Tilgner (1962), and

138

a pollen analysis with the use of method established by the International Commission of Bee

139

Botany (Louveaux Maurizio & Vorwohl, 1978). The following parameters of chosen honeys

140

were determined:

141

1) water content by measurement of refractive index with a use of Atago RX-5000i

143 144

EP

refractometer (AOAC, 1995; Anonymous, 2001);

AC C

142

TE D

136

2) total ash content by incinerating honey samples in a muffle furnace at a temperature of 550°C;

145

3) reducing sugars, total sugars and sucrose content using Agilent 1200 series HPLC along

146

with RI detector. 20 µL of each sample was injected onto an Agilent Hi-Plex Ca, 7.7 × 300

147

mm, 8 µm column at 85 °C with a flow rate of 0.6 mL/min. Pure water was used as eluent;

6

ACCEPTED MANUSCRIPT

149 150 151 152 153

4) active acidity (pH) of aqueous 20 g/100g solutions using a CX-721 multi-function computer measuring instrument; 5) total acidity of aqueous 20 g/100g solutions; acidic honey components were neutralized by a standard solution of sodium hydroxide; 6) specific electrical conductivity of aqueous 20 g/100g solutions using a CX-721 multi-

RI PT

148

function computer measuring instrument;

7) dynamic viscosity of aqueous 20 g/100g solutions using a Ubbelohde viscometer TC SI

155

Analytics; a honey solution flow in a capillary of Ubbelohde viscometer was measured

156

(PN-87/C-8929/20, 1987);

SC

154

8) diastatic number using photometric method with insoluble starch conjugated with blue dye

158

as a substrate. Amylase hydrolyzes starch into water-soluble fragments forming joints with

159

blue dye, which absorbance is measured by spectrophotometry using a Spectrophotometer

160

V5600 Vis at a wavelength of 620 nm;

M AN U

157

9) content of 5-HMF using HPLC Shimadzu Prominence with UV detector. 20 µL of each

162

sample was injected onto a RP-18 column (250x4 mm, 5 µm particle diameter) with a flow

163

rate of 1 mL/min. A mixture of water (1 g/100g acetic acid) : methanol (90:10 mL:mL)

164

was used as eluent. Detection was performed at 285 nm. The solution before application to

165

the column was filtered through a filter of a 0.45 µm diameter.

EP

TE D

161

10) proline content using U-2001 Hitachi Instruments Inc. Spectrophotometer. Proline was

167

isolated from other amino acids with isopropanol (water solution 1:1 (mL:mL)). To enable

168

colorimetric measurement colored complex with 3 g/100g solution of ninhydrin in

169

dimethoxyethanol (g:g) in an environment of concentrated formic acid was prepared.

170

Spectrophotometric detection was performed at 520 nm.

171

AC C

166

All samples were analyzed in triplicate.

7

ACCEPTED MANUSCRIPT The obtained results were analyzed with methods of descriptive statistics. To develop

173

a model for identification of type and variety of honey calculations with the use of a data

174

mining method C&RT (Classification and Regression Trees) were made. This is the method

175

that is applied to construct prediction and descriptive models. The model is constructed during

176

a process of recursively dividing a set of observations into disjoint sub-sets. A graphic

177

representation of the model is the so-called Tree. The Classification Trees are employed in a

178

situation, where a dependent variable is measured on a nominal or ordinal scale, and the

179

Regression Trees there, where the scale of a dependent variable is at least an interval scale.

180

The model is being built for the purpose of obtaining maximally homogenous sub-sets from

181

the point of view of the value of the dependent variable. This is a multi-stage process and a

182

different independent variable can be utilized at every consecutive stage since all predictors

183

are analyzed at every stage and such a predictor is selected that provides the best division of

184

the node, i.e. a predictor that produces the maximally homogeneous subsets (Hastie,

185

Tibshirani & Friedman, 2009). STATISTICA version 10 software was used for all statistical

186

analyses.

187

3. Results and discussion

EP

188 189

The results of physicochemical analyses of all varietal honey samples were presented

AC C

190

TE D

M AN U

SC

RI PT

172

191

in Table 1. In Table 2 the results of physicochemical analyses of individual varieties/types of

192

honey were shown. Based on these data, high level of the quality of analyzed honeys was

193

demonstrated. Typicality (Anonymous, 2001; Anonymous, 1974; Anonymous, 2004) of all

194

samples was confirmed. This proved suitability of data concerning all tested honeys for

195

modelling.

8

ACCEPTED MANUSCRIPT The Classification Model was constructed with help C&RT classification trees to

197

categorize tested honey samples into an appropriate category (one of the eight categories

198

indicated). The dependent variable was a varying Class (type/variety). All the analyzed

199

physicochemical features of honey were applied as predictors. A C&RT algorithm with equal

200

misclassification costs was selected. A classification tree was constructed using the Gini

201

measure of node impurity and minimal deviance-complexity pruning method was applied. A

202

minimum size of the node being divided was assumed at a putative level of 10% (later, it

203

occurred that this criterion had no impact on the model construction process).

SC

RI PT

196

At first, based on the above described settings a model as presented on the Fig. 1 was

205

built. Next, a v-fold cross-validation was applied (v=10). The application of the v-fold cross-

206

validation aimed at improving the generalization ability of the model under construction (thus,

207

at reducing the complexity thereof). The model built upon the application of the v-fold cross-

208

validation was deemed to be the standard model and its graphic representation is given in Fig.

209

2.

211

TE D

210

M AN U

204

Based on the Decision Tree developed, the following rules were set (consecutively from the left to the right):

1. if the value of specific electrical conductivity is lower than or equal to 4.50 x 10(-4)

213

S/cm, the sucrose content is lower than or equal to 3.105 g/100g, and the content of

214

reducing sugars is lower than or equal to 78.585 g/100g, then the honey identified is

AC C

215

EP

212

from Class G (buckwheat variety of honey);

216

2. if the value of specific electrical conductivity is lower than or equal to 4.50 x 10(-4)

217

S/cm-, the sucrose content is lower than or equal to 3.105 g/100g, and the content of

218

reducing sugars is higher than 78.585 g/100g, then the honey identified is from Class

219

R (rape variety of honey);

9

ACCEPTED MANUSCRIPT 220

3. if the value of specific electrical conductivity is lower than or equal to 4.50 x 10(-4)

221

S/cm and the sucrose content is higher 3.105 g/100g, then the honey identified is from

222

Class A (acacia variety of honey); 4. if the value of specific electrical conductivity is higher than 4.50 x 10(-4) S/cm, the

224

total ash content is lower than or equal to 0.3953 g/100g, the content of sucrose is

225

lower than or equal to 1.30 g/100g, and the content of reducing sugars is lower than or

226

equal to 73.92 g/100g, then the honey identified is from Class WR (heather variety of

227

honey);

SC

RI PT

223

5. if the value of specific electrical conductivity is higher than 4.50 x 10(-4) S/cm, the

229

total ash content is lower than or equal to 0.3953 g/100g, the content of sucrose is

230

lower than or equal to 1.3 g/100g, and the content of reducing sugars is higher 73.92

231

g/100g, then the honey identified is from Class L (linden variety of honey);

M AN U

228

6. if the value of specific electrical conductivity is higher than 4.50 x 10(-4)S/cm, the

233

total ash content is lower than or equal to 0.3953 g/100g, and the content of sucrose is

234

higher than 1.3005 g/100g, then the honey identified is from Class W (multifloral

235

variety of honey);

TE D

232

7. if the value of specific electrical conductivity is higher than 4.50 x 10(-4) S/cm, the

237

total ash content is higher than 0.3953 g/100g, and the total acidity is lower than or

238

equal to 2.80°, then the honey identified are from Class NS (nectar-honeydew type of

AC C

239

EP

236

honey);

240

8. if the value of specific electrical conductivity is higher than 4.50 x 10(-4) S/cm, the

241

total ash content is higher than 0.3953 g/100g, and the total acidity is higher than

242

2.80°, then the honey identified is from Class S (honeydew type of honey).

243

10

ACCEPTED MANUSCRIPT 244 245

To check if the model reproduce the structure of dataset the classification matrix with the v-fold cross-validation applied was constructed (Table 3). The results prove that the model developed was mistaken in one case only. The honey

247

sample of heather variety was incorrectly classified as a multifloral honey type. Therefore, the

248

accuracy of the model is: 98.61%. The developed model can successfully be used for

249

confirmation of the botanical origin of honey.

RI PT

246

A thorough analysis of the available literature showed that there are no articles

251

presenting sufficient data that would enable verification of the usefulness of the proposed

252

model for honeys produced in other parts of the world. Nevertheless, the findings of Sahinler,

253

Sahinler & Gul (2009) confirm the advisability of using chemometric analysis to verify

254

botanical origin of honey. . They managed to differentiate 50 honey samples from Turkey, by

255

their botanical origin, using discriminant analysis to the following physicochemical

256

parameters: mineral content, moisture content, pH, acidity, sugar composition (invert sugar,

257

sucrose), diastase activity and hydroxymethylfurfural.

258

260

4. Conclusions

EP

259

TE D

M AN U

SC

250

Determination of the level of some individual, honey quality parameters, and even the

262

compilation thereof, does not result in obtaining definite information on the botanical origin

263

of honey. Therefore as a result of this study a new classification model was constructed. Clear

264

rules that characterize every type/variety of honey were set. The developed model reproduces

265

the structure of dataset very well. Its accuracy is 98.61%.

AC C

261

266

11

ACCEPTED MANUSCRIPT 267

Funding: The research was subsidized by the Ministry of Science and Higher Education -

268

grant for the maintenance of the research potential, awarded to the Faculty of Commodity

269

Science of the Cracow University of Economics.

270

References

RI PT

271 272 273

Aliferis,

K. A., Tarantilis, P. A., Harizanis, P. C., & Alissandakis, E. (2010). Botanical

discrimination and classification of honey samples applying gas chromatography/mass

275

spectrometry fingerprinting of headspace volatile compounds. Food Chemistry, 121(3), 856-

276

862.

M AN U

SC

274

277

Anonymous (2004). Regulation of the Polish Minister of Agriculture and Rural Development

278

amending regulation on detailed requirements for commercial quality of honey, Dz. U. 40,

279

poz.370, Warszawa, Poland. .

281

Anonymous (2001). Codex Standard for Honey. European Regional Standard 12, Rev. 2., Rome, Italy.

TE D

280

Anonymous (1974). Council Directive on the harmonization of the Member States relating to

283

honey (74/409/EEC), Brussels, Belgium. AOAC (1995). Official Methods of Analysis.

284

Association of Official Analytical Chemists, Washington, USA.

286 287 288

AC C

285

EP

282

Bogdanov, S., Gallmann, P. (2008). Authenticity of honey and other bee products. State of the art. ALP Science, 520, 1-12. Castro, R. M., Escamilla, M. J., & Reig, F. B. (1992). Evaluation of the color of some unifloral honey types as a characterization parameter. Journal of AOAC International, 75, 537-542.

289

Castro-Vázquez, L., Leon-Ruiz, V., Alañon, M. E., Pérez-Coello, M. S., & González-Porto, A.

290

V. (2014). Floral origin markers for authenticating Lavandin honey (Lavandula angustifolia x

12

ACCEPTED MANUSCRIPT 291

latifolia). Discrimination from Lavender honey (Lavandula latifolia), Food Control, 37, 362-

292

370. Chen, L., Wang, J., Ye, Z., Zhao, J., Xue, X., Heyden, Y. V., & Sun, Q. (2012). Classification of

294

Chinese honeys according to their floral origin by near infrared spectroscopy. Food

295

Chemistry, 135, 338–342.

RI PT

293

Di Bella, G., Lo Turco, V., Potorti, A. G., Bua, G. D., Fede, M. R., & Dugo G. (2015).

297

Geographical discrimination of Italian honey by multi-element analysis with a chemometric

298

approach, Journal of Food Composition and Analysis, 44, 25-35.

SC

296

Escriche, I., Kadar, M., Juan-Borras, M., & Domenech, E. (2011). Using flavonoids, phenolic

300

compounds and headspace volatile profile for botanical authentication of lemon and orange

301

honeys. Food Research International, 44, 1504-1513.

M AN U

299

Escuredo, O., González-Martín, M. I., Rodríguez-Flores, M. S., & Seijo, M. C. (2015). Near

303

infrared spectroscopy applied to the rapid prediction of the floral origin and mineral content

304

of honeys. Food Chemistry, 170, 47-54.

TE D

302

Gašić, U. M., Natić, M. M., Mišić, D. M., Lušić, D. V., Milojković-Opsenica,D. M., Tešić, Ž. L.,

306

& Lušić D. (2015). Chemical markers for the authentication of unifloral Salvia officinalis L.

307

honey. Journal of Food Composition and Analysis, 44, 128-138.

308

EP

305

Hastie, T., Tibshirani R., & Friedman, J. (2009). The elements of Statistical Learning: Data Mining Inference and Prediction, http://www-stat.stanford.edu~tibs/ElemStatlearn/ Accessed

310

20.01.16.

AC C

309

311

Karabagias, I. K., Badeka, A. V., Kontakos, S., Karabournioti, S., & Kontominas, M. G. (2014).

312

Botanical discrimination of Greek unifloral honeys with physico-chemical and chemometric

313

analyses. Food Chemistry, 165, 181-190.

314

Karabagias, I. K., Vavoura, M. V., Nikolaou, C., Badeka, A. V., Kontakos, S. & Kontominas, M.

315

G. (2014). Floral authentication of Greek unifloral honeys based on the combination of

13

ACCEPTED MANUSCRIPT 316

phenolic compounds, physicochemical parameters and chemometrics. Food Research

317

International, 62, 753-760. Kelly, S. D. (2003). Using stable isotope ratio mass spectrometry (IRMS) in food authentication

319

and traceability. In M. Lees (Ed.), Food authenticity and traceability (pp. 156–183).

320

Cambridge, England: Woodhead Publishing Ltd.

RI PT

318

Kerlvliet, D. (1992). De bepaling van de botanische herkomst van honig d. m. r. organoleptische

322

eigenschappen, de pH, de elektrische geleiding en de mikroskopische eigenschappen.

323

Inspectie Gezondheidsbescherming Keuringsdienst van Waren, 22(4), 208-227.

325

Louveaux, J., Maurizio, A., & Vorwohl, G. (1978). Methods of melissopalynology (republished and updated). Bee World, 59, 139-157.

M AN U

324

SC

321

326

Meda, A., Lamiec, Ch. E., Romito, M., Millogo, J., & Nacoulma, O. (2005). Determination of

327

the total phenolic, flavonoid and proline contents in Burkina Fasan honey, as well as their

328

radical scavenging activity. Food Chemistry, 91(3), 571-577.

330

Ojeda de Rodrı́guez, G., Sulbarán de Ferrer, B., Ferre, A., & Rodrı́guez, B. (2004).

TE D

329

Characterization of honey produced in Venezuela. Food Chemistry, 84(4), 499-502. Persano Oddo, L., Piana, L., Bogdanov, S., Bentabol, A., Gotsiu, P., Kerkvliet, J., Martin, P.,

332

Morlot, M., Valbuena, A. O., Ruoff, K., & Von der Ohe, K. (2004). Botanical species giving

333

unifloral honey in Europe. Apidologie, 35 (Special Issue), 82-93.

335 336 337

PN-87/C-8929/20. (1987). Polish Standard: Determination of viscosity number in dilute solution

AC C

334

EP

331

by the Ubbelohde viscosimeter. Popek, S. (1998). Electrical conductivity as an indicator of the quality of nectar honeys. Forum Ware, 1-4, 75-70.

338

Ruoff, K., Luginbühl, W., Kilchenmann, V., Bosset, J. O., von der Ohe, K., von der Ohe, W., &

339

Amadó, R. (2007). Authentication of the botanical origin of honey using profiles of classical

340

measurands and discriminant analysis. Apidologie, 38, 438-452.

14

ACCEPTED MANUSCRIPT 341 342 343 344

Sahinler, S., Sahinler, N., & Gul, A. (2009). Determination of honey botanical origin by using discriminant analysis. Journal of Animal and Veterinary Advances, 8(3), 488-491. Scandurra, G., Tripodi, G., & Verzera, A. (2013). Impedance spectroscopy for rapid determination of honey floral origin. Journal of Food Engineering, 119, 738-743. Serrano, S., Villareho, M., Espejo, R., & Jodral, M. (2004). Chemical and physical parameters of

346

Andalusian honey: classification of Citrus and Eucalyptus honey by discriminant analysis.

347

Food Chemistry, 87(4), 619-625.

349

Soares, S., Amaral, J. S., Oliveira, M. B., & Mafra, I. (2015). Improving DNA isolation from

SC

348

RI PT

345

honey for the botanical origin identification. Food Control, 48, 130-136.

Špánik, I., Pažitná, A., Šiška, P. & Szolcsányi P. (2014). The determination of botanical origin of

351

honeys based on enantiomer distribution of chiral volatile organic compounds. Food

352

Chemistry, 158, 497-503.

M AN U

350

Spiteri, M., Jamin, E., Thomas, F., Rebours, A., Lees, M., Rogers, K. M., & Rutledge, D. N.

354

(2015). Fast and global authenticity screening of honey using 1H-NMR profiling. Food

355

Chemistry, 189, 60-66.

TE D

353

Stephens, J. M., Schlothauer, R. C., Morris, B. D., Yang, D., Fearnley, L., Greenwood, D. R., &

357

Loomes, K. M. (2010). Phenolic compounds and methylglyoxal in some New Zealand

358

Manuka and Kanuka honeys. Food Chemistry, 120, 78-86.

EP

356

Terrab, A., González, A.G., Díez, M. J., & Heredia, F. J. (2003). Characterisation of Moroccan

360

unifloral honey using multivariate analysis. European Food Research and Technology,

361

218(1), 88-95.

AC C

359

362

Tewari, J., & Irudayaraj, J. (2004). Quantification of saccharides in multiple floral honey using

363

Fourier transform infrared microattenuated total reflectance spectroscopy. Journal of

364

Agricultural and Food Chemistry, 52, 3237–3243.

365

Tilgner, D. J. (1962). Dilution Test for Odour and Flavour Analysis. Food Technology, 2, 58-64.

15

ACCEPTED MANUSCRIPT Wei, Z. & Wang J. (2014). Tracing floral and geographical origins of honeys by potentiometric

367

and voltammetric electronic tongue. Computers and Electronics in Agriculture, 108, 112-122.

368

Yücel, Y. & Sultanoglu, P. (2013). Characterization of Hatay honeys according to their multi–

369

element analysis using ICP–OES combined with chemometrics. Food Chemistry, 140, 231-

370

237.

RI PT

366

Zhao, J., Dua, X., Cheng, N., Chen, L., Xue, X., Zhao, J., Wuc, L., & Cao, W. (2016).

372

Identification of monofloral honeys using HPLC–ECD and chemometrics. Food Chemistry,

373

194, 167-174.

AC C

EP

TE D

M AN U

SC

371

16

ACCEPTED MANUSCRIPT Table 1. Results of physicochemical parameters of all honey samples and descriptive

Measure

n

Mean

Median

Min.

Max.

SD

Electrical conductivity

72

5.97

6.01

1.66

10.88

2.72

Total ash (g/100g)

72

0.2905

0.2212

0.0887

0.7123

0.0923

Extract (g/100g)

72

81.91

82.15

78.00

84.30

1.91

Water (g/100g)

72

16.53

16.25

14.10

20.20

1.92

Acidity (°)

72

2.12

2.00

1.10

4.10

0.72

Total sugar (g/100g)

72

76.48

75.22

68.67

87.36

4.83

Reducing sugars (g/100g)

72

73.86

73.15

66.12

84.21

4.52

Sucrose (g/100g)

72

2.36

1.63

0.36

7.68

1.85

Dynamic viscosity, (mPa•s )

72

1.64

1.61

SC

statistics.

1.11

2.23

0.15

pH

72

3.85

3.84

3.39

4.71

0.24

Diastatic number

72

16.25

13.90

6.50

38.50

6.02

5-HMF (mg/kg)

72

1.62

1.11

0.25

4.54

1.21

Proline (mg/kg)

72

47.51

45.00

26.20

90.50

4.71

M AN U

TE D EP AC C

RI PT

(x 10(-4) S/cm)

ACCEPTED MANUSCRIPT Table 2. Results of physicochemical parameters of individual types/varieties of honey Group of honey

Acacia

Linden

Multi- Buckwheat Heather

Rape

Honeydew

floral 9

9

9

honeydew 9

9

9

9

9

x ± SD

Measurement

RI PT

n

Nectar-

Electrical

2.33±

5.61

6.79

3.63

6.06

3.51±

9.64

10.19

conductivity (x

0.38

± 0.27

± 0.36

± 0.32

± 0.23

0.24

± 0.43

± 0.58

Total ash

0.1011

0.1703

0.2791

0.2502

0.2111

0.1312

0.6314

0.5727

(g/100g)

± 0.02

± 0.03

± 0.02

± 0.03

± 0.03

± 0.02

± 0.04

± 0.05

Extract (g/100g)

81.31

80.79

82.82

82.84

80.17

81.36

82.79

83.19

± 1.78

± 2.03

± 2.01

± 1.73

± 1.63

± 2.03

± 1.45

± 1.61

17.17

17.71

15.61

± 1.78

± 2.03

± 2.01

1.36

2.03

1.82

± 0.21

± 0.29

± 0.19

Total sugar

83.90

80.90

75.00

(g/100g)

± 2,54

± 3.23

± 1.25

Reducing sugars

77.03

79.43

73.291

(g/100g)

± 3.50

± 3.25

Sucrose (g/100g)

6.03 ± 0.94

M AN U 18.37

16.90

15.66

15.23

± 1.73

± 1.63

± 2.03

± 1.45

± 1.61

2.51

2.30

1.46

1.80

3.68

± 0.18

± 0.19

± 0.11

± 0.27

± 0.30

73.60

72.90

80.70

72.19

72.70

± 2.21

± 1.23

± 1.74

± 1.85

± 2.21

72.15

71.82

79.42±

69.22

68.53

±1.23

± 2.20

± 1.21

1.78

± 1.82

± 2.23

0.72

2.19

1.40

0.94

1.13

2.49

4.00

± 0.25

± 0.69

± 0.31

± 0.28

± 0.38

± 0.33

± 0.62

TE D

Acidity (°)

15.69

EP

Water (g/100g)

SC

10(-4) S/cm)

1.755

1.574

1.5432

1.691

1.684

1.745

1.548

1.615

)

± 0.02

± 0.07

± 0.08

± 0.08

± 0.09

± 0.11

± 0.12

± 0.11

pH

3.76

3.69

3.78

3.71

3.97

3.66

4.02

4.18

± 0.12

± 0.21

± 0.15

± 0.11

± 0.09

± 0.11

± 0.10

± 0.16

9.92

14.99

15.83

21.18

14.88

11.84

19.91

21.44

± 2.64

± 3.87

± 4.21

± 3.11

± 5.27

± 4.65

± 5.23

± 5.73

0.68

0.95

0.97

1.72

0.92

0.86

3.50

3.38

± 0.14

± 0.16

± 0.21

± 0.27

± 0.15

± 0.25

± 0.53

± 0.40

42.80

46.17

38.87

53.33

37.78

34.01

61.68

65.47

± 5.56

± 6.08

± 4.21

± 6.44

± 3.96

± 6.02

± 7.47

± 6.85

AC C

Viscosity (mPa•s

Diastatic number

5-HMF (mg/kg)

Proline (mg/kg)

ACCEPTED MANUSCRIPT Table 3. Classification matrix for the model with v-fold cross-validation applied N/A

P/L

N/F

N/B

N/H

N/R

N / HD

N / HDH

9 9

P/F

9

1

P/B

9

P/H

8

P/R

RI PT

P/A

N/L

9

9

SC

P / HD P / HDH

P / types/varieties of honey - predicted number of observations

AC C

EP

TE D

M AN U

N / types/varieties of honey - number of observations in the data set

9

AC C

EP

TE D

M AN U

SC

RI PT

ACCEPTED MANUSCRIPT

Fig. 1. Model in the form of the Decision Tree. A , Acacia honey; L , Linden honey; W, Multifloral honey; G, Buckwheat honey; WR, Heather honey; R, Rape honey; NS, Nectar-honeydew honey; S, Honeydew honey

AC C

EP

TE D

M AN U

SC

RI PT

ACCEPTED MANUSCRIPT

Fig. 2. Model in the form of the Decision Tree upon the application of v-fold cross-validation. A , Acacia honey; L , Linden honey; W, Multifloral honey; G, Buckwheat honey; WR, Heather honey; R, Rape honey; NS, Nectar-honeydew honey; S, Honeydew honey

ACCEPTED MANUSCRIPT A new model to identify botanical origin of Polish honeys based on the physicochemical parameters and chemometric analysis

Model to identify biological origin of varietal honey was created.

RI PT

Research Highlights

Data mining method C&RT proved useful for honey’s origin verification model creation. Physicochemical parameters of honey samples were determined.

AC C

EP

TE D

M AN U

SC

Quality of varietal honeys’ samples was evaluated.