U. KROPF et al.: Linear Regression Model for Slovenian Honey, Food Technol. Biotechnol.

46 (3) 335–340 (2008) 335

ISSN 1330-9862 scientific note

(FTB-1809)

Linear Regression Model of the Ash Mass Fraction

and Electrical Conductivity for Slovenian Honey

Ur{ka Kropf*, Mojca Jamnik, Jasna Bertoncelj and Terezija Golob

Department of Food Science and Technology, Biotechnical Faculty,
University of Ljubljana, Jamnikarjeva 101, SI-1000 Ljubljana, Slovenia
Received: November 10, 2006
Revised version: November 9, 2007
Accepted: November 14, 2007

Summary
Mass fraction of ash is a quality criterion for determining the botanical origin of ho-
ney. At present, this parameter is generally being replaced by the measurement of electri-
cal conductivity (k). The value k depends on the ash and acid content of honey; the higher
their content, the higher the resulting conductivity. A linear regression model for the rela-
tionship between ash and electrical conductivity has been established for Slovenian honey
by analysing 290 samples of Slovenian honey (including acacia, lime, chestnut, spruce, fir,
multifloral and mixed forest honeydew honey). The obtained model differs from the one
proposed by the International Honey Commission (IHC) in the slope, but not in the sec-
tion part of the relation formula. Therefore, the Slovenian model is recommended when
calculating the ash mass fraction from the results of electrical conductivity in samples of
Slovenian honey.

Key words: honey, ash, electrical conductivity, linear regression model

Electrical conductivity depends on the ash and acid

Introduction content of honey, the higher their content, the higher the
resulting conductivity (2). The relation between k and
In the present research, which is part of an exten-
mineral content has already been demonstrated by many
sive study of the physicochemical and sensory charac-
researchers, who have determined that the above-men-
teristics of Slovenian unifloral and mixed honeys, the
tioned parameters are in a linear relationship, which can
behaviour of five unifloral and two mixed types of hon-
be expressed with various equations (4,8–12). A higher
ey is examined with respect to the electrical conductiv-
ity and ash mass fraction. mineral content results in a higher k value. Minerals are
introduced into honey primarily with the pollen. The
Electrical conductivity and ash mass fraction are mineral and ash mass fraction therefore depend on the
measures of the mineral content of honey (1–4). The ash predominant pollen present in honey (2); they are, as
mass fraction is a quality criterion for determining the such, appropriate parameters for honey type determina-
botanical origin of honey (2,4–6); brighter honeys usu- tion (4,8–12).
ally containing fewer elements than darker ones. At pre-
sent, the determination of ash mass fraction is generally The pollen in honey derives from both nectar and
being replaced by the measurement of electrical conduc- nectarless plants. When calculating the percentage of
tivity (k) (2–4,7). The total ash mass fraction could be re- pollen present during pollen analysis, the pollen of nec-
tained as a quality factor during a transition period until tarless plants is excluded. In this way only pollen from
conductivity is accepted as a worldwide standard (7). nectar plants determines the type of honey in pollen ana-

*Corresponding author; Phone: ++386 1 42 31 161; Fax: ++386 1 25 66 296; E-mail: [email protected]
336 U. KROPF et al.: Linear Regression Model for Slovenian Honey, Food Technol. Biotechnol. 46 (3) 335–340 (2008)

lysis. But the ash mass fraction and k value depend on origin, forest honeydew honey (mixed forest trees), fir
the whole pollen spectrum of the honey. In spite of this, (Abies alba) and spruce honey (Picea abies) of honeydew
mineral and ash mass fraction are honey type depend- origin. Among these honey types, great differences exist,
ent (10). both in k and ash mass fraction as well as in the sensory
A linear relationship exists between electrical con- characteristics and pollen spectrum.
ductivity on the one side and soluble and insoluble ash
(11), sulphated ash (9) and total ash mass fraction (4,8,
Materials and Methods
9,12) on the other side. Italian model (4,8) that describes
a relationship between electrical conductivity and total All samples (290) included in this research were
ash mass fraction is: k=1.74 w(ash)+0.14; where k is the produced in the years 2004 and 2005 and were gathered
electrical conductivity in mS/cm, and w(ash) is the total directly from Slovenian beekeepers. The sampling was
ash mass fraction in g per 100 g of honey. This model is carried out by beekeepers immediately after harvesting,
a general one; it covers a range of k from 0.09 to 2.11 and in accordance with written instructions. Honey
mS/cm and, according to Vorwohl, it has been proposed samples were stored at room temperature in plastic con-
for use around the whole Europe by International Hon- tainers. In this way no mineral or dust contamination
ey Commission (IHC) (1). This model should replace an could occur. Collected samples were produced in differ-
older, slower and more difficult method for determina- ent regions of Slovenia. Samples were of different botan-
tion of total ash mass fraction by ashing. The main prob- ical origin, mostly of mixed botanical origin (185 sam-
lem, however, is that this model is based on Italian hon- ples), with no specific sensory characteristics of a partic-
ey types and has not been tested in many other honey- ular honey type. Other samples (105) included 15 sam-
-producing countries in Europe. The other weakness of ples each of the seven main Slovenian honey types (aca-
this model is that it cannot be used for samples with k cia, multifloral, lime, chestnut, forest, spruce and fir).
equal to or lower than 0.14 mS/cm. Nevertheless, in It-
aly and other parts of southern Europe there are many Sample preparation
honey types, such as rosemary and citrus, which possess
Ash mass fraction and electrical conductivity were
very low k, usually below 0.14 mS/cm.
determined in 290 samples of Slovenian honey accord-
Spanish model for determining electrical conductiv- ing to the AOAC method (14) and the harmonised me-
ity, based on Basque honey (9), is: k= 1.205 w(ash)+0.092. thods of the International Honey Commission (7). Before
This model was constructed by using data for k measured the analyses, all crystallised samples were homogenised
in honey solution prepared according to the procedure and liquefied at 40 °C in order to enable water content
proposed by Sancho et al. (13), where 10 g of honey is determination.
weighed per 75 mL of distilled water. This procedure
does not need water content information before electri- Methods
cal conductivity determination. Therefore, this model is
not directly comparable to the Italian model, but it has Content of water
to be recalculated first. Spanish researchers who set up The refractive method is based on the principle that
this model performed recalculation and found out that the refractive index of honey increases with the solids
both models are similar. Sancho’s procedure is very sim- content. Water content of honey is determined from the
ple, but unified procedures in honey analysis should be refractive index of the honey by reference to a standard
used around Europe to enable researchers to compare table. The method is recommended in harmonised meth-
the results without any recalculations needed because of ods (7).
different honey solution preparations. Therefore, in our
Electrical conductivity (Method A)
opinion, the IHC procedures should be followed as much
as possible. This method is valid for the determination of elec-
trical conductivity (k) of honey in the range of 0.1–3
Another example from the literature is the Irish mo-
mS/cm. Electrical conductivity is measured using a 20
del. Downey et al. (12) determined a linear regression of
% (by mass per volume) solution in water, where the 20
the total ash mass fraction and electrical conductivity to
% refers to the honey dry matter. This method is recom-
be k=0.771 w(ash)+0.1084 for Irish honey. This model is
mended in the harmonised methods (7).
accurate only for low k values (0.1–0.5 mS/cm) and
therefore is not comparable to the Italian or Spanish one. Electrical conductivity (Method B)
Europe or countries of the European Union are very di- A modified method was applied for the determina-
verse in climate and pedology conditions; thus honeys tion of electrical conductivity (k) in honey using a 20 %
produced in different parts are variable. It is therefore (by mass) honey solution in water, where the 20 % refers
unreasonable to expect honey from Europe to be homo- to the honey dry matter. This method is less time and
geneous and that a model established for honey from a glassware consuming. We used it to assess the differen-
distinct part of Europe would be accurate for the whole ces between the two methods.
of Europe. Consequently, the question arises, which mo-
del should be used for Slovenian honey? Ash mass fraction
There are 7 main types of honey in Slovenia: acacia Ash mass fraction of honey is the residue obtained
(Robinia pseudoacacia) and multifloral (mixed meadow by a defined procedure and is expressed as a percentage
plants and fruit trees) of nectar origin, lime (Tilia spp.) by mass. The honey is ashed at 600 oC and the residue is
and chestnut (Castanea sativa) of nectar and honeydew weighed (14,15).
U. KROPF et al.: Linear Regression Model for Slovenian Honey, Food Technol. Biotechnol. 46 (3) 335–340 (2008) 337

Sensory analysis and lime. There was also no difference among forest,
Sensory analysis of all samples was used for select- spruce and fir honey, but these types differed from all
ing the samples with characteristic properties of each the others.
type of Slovenian honey. Sensory characteristics describ- Water content completely depends on the maturity
ed by Golob et al. (16) for each type of Slovenian honey of honey; it was therefore not meaningful for compari-
were used in this analysis. Sensory analysis includes ap- son among the different types of honey.
pearance, smell, taste, and aroma assessment. Through The average mass fraction of ash in our acacia hon-
sensory analysis, 15 most representative samples of each ey was lower than in Croatian honey (17). On the other
type were selected from the 290 samples. All samples hand, the average mass fraction of ash in Slovenian mul-
were sensory assessed by at least three members of the tifloral honey was higher than in Croatian multifloral
sensory panel. honey. The range of ash mass fraction in our samples,
from 0.04 to 1.04 g per 100 g, was wider than previously
Statistical analysis
reported for Spanish thyme honey (18), multifloral hon-
All analyses were carried out in triplicate. For calcu- ey (5,19) and chestnut honey (5). The average ash mass
lating the relationship between ash mass fraction and k, fraction in their samples was higher than we measured
only average values of each sample were used. The fol- for nectar honey types, whereas the average k was low-
lowing statistical parameters were calculated: average er for thyme honey than for our floral and lime honey
value, standard deviation and coefficient of variability. and higher than k for acacia honey. The average mass
After the correlation analysis, analysis of variance fraction of ash in Slovenian multifloral honey was com-
(ANOVA) and Duncan’s test were performed. Correla- parable to the ash mass fraction determined for Indian
tions were obtained by Pearson’s correlation coefficient multifloral honey (20).
in bivariate linear correlations. Differences between means
Our main prediction about linear regression model
at the 95 % (p£0.05) confidence level were considered
for k and ash mass fraction was that different models
significant.
were valid only for specific types of honey. However, as
can be seen in the last two columns of Table 1, this antic-
Results and Discussion ipation was incorrect, since no statistically significant
differences between models for different types of honey
Fifteen samples of 7 types of honey (in total 105 had been found. The linear regression model of ash mass
samples) were analysed for water content, ash mass frac- fraction and k is therefore independent of honey type.
tion and k. The average values and standard deviations From the available literature we understand that this
(SD) are presented in Table 1. Using ANOVA and Dun- hypothesis has not previously been tested.
can’s test, we determined that the average values for ash All Slovenian honey types can be described with
mass fraction and k were statistically significantly differ- one general model, which is presented in Fig. 1. Yet, this
ent among some types of Slovenian honey at p£0.05. model is imperfect due to the lack of data for k in the
Acacia and chestnut honey differed from all other types, ranges from 0.8 to 1.0 mS/cm and from 0.25 to 0.5 mS/
whereas no difference existed between the multifloral cm, as well as above 1.5 mS/cm.

Table 1. Average values and SD for water content, ash mass fraction and kB

Linear regression between k

Number
Type of honey Basic statistics w(water)/% w(ash)/% kB/(mS/cm) and w(ash)
of samples
Model R2
Average±SD 16.2±1.0 (0.06±0.01)a (0.20±0.03)a
Acacia 15 k=2.09 w(ash)+0.08 0.81
Range 14.5–15.5 0.04–0.09 0.16–0.26
Average±SD 16.0±0.8 (0.21±0.05)b (0.58±0.12)b
Multifloral 15 k=2.19 w(ash)+0.12 0.86
Range 15.2–17.8 0.11–0.27 0.33–0.78
Average±SD 16.7±1.0 (0.25±0.04)b (0.73±0.07)b
Lime 15 k=1.95 w(ash)+0.23 0.82
Range 15.2–18.9 0.18–0.30 0.55–0.86
Average±SD 16.2±1.1 (0.70±0.13)d (1.62±0.28)d
Chestnut 15 k=2.07 w(ash)+0.15 0.92
Range 14.6–18.0 0.55–1.04 1.20–2.25
Average±SD 14.9±0.6 (0.53±0.06)c (1.28±0.13)c
Forest 15 k=1.91 w(ash) +0.26 0.89
Range 14.1–15.8 0.44–0.63 1.06–1.47
Average±SD 15.8±1.1 (0.57±0.09)c (1.31±0.19)c
Spruce 15 k=1.93 w(ash)+0.22 0.88
Range 14.3–18.4 0.41–0.71 0.96–1.47
Average±SD 15.3±0.8 (0.55±0.08)c (1.29±0.15)c
Fir 15 k=1.99 w(ash)+0.21 0.98
Range 13.8–16.6 0.38–0.65 1.07–1.63

Data with different superscript in the same column are statistically significantly different at p£0.05
 338 U. KROPF et al.: Linear Regression Model for Slovenian Honey, Food Technol. Biotechnol. 46 (3) 335–340 (2008)

2.50 these two methods. Method B gave results that were on

average 3 % lower than the results of method A. A sta-
2.00
tistically significant relationship (R2=0.999) exists be-
kB /(mS/cm)

1.50
tween these two methods. We prefer the method B due
to savings in time and glassware use, but only if used
1.00 with proper correction. Our correction and relationship
between the two methods is presented in Fig. 3, where
kB =2.04 w(ash)+0.12
0.50 kB is on the abscissa while kA is on the ordinate.
R 2=0.99
0.00
2.50
0.00 0.20 0.40 0.60 0.80 1.00 1.20
w(ash)/% 2.00

kB /(mS/cm)
Fig 1. Linear regression of kB and w(ash) in 105 samples (7
1.50
types and 15 samples of each type)
1.00

Consequently, we decided to expand this research kB =1.038 kA –0.007

0.50
and include more honey samples. The number of sam- R 2=0.999
ples thus rose from 105 to 290. These extra samples were 0.00
of mixed botanical origin and could not be classified as 0.00 0.50 1.00 1.50 2.00 2.50
just a single type of honey. As such, these samples were kA /(mS/cm)
only used for linear regression model construction. The
Fig 3. Linear relationship between two methods for k determi-
model presented in Fig. 2 was generated from the data
nation: kB and kA
for k and ash mass fraction of all 290 samples. Here, the
k varies from 0.16 to 2.25 mS/cm and the ash mass frac-
tion varies from 0.04 to 1.04 g per 100 g of honey. The All our results were recalculated using the above-
determined linear regression model is therefore valid -mentioned relationship between electrical conductivity
only in this range. measured by the two methods and a new model was
constructed from the new data. This model is presented
in Fig. 4.
2.50
2.50
2.00
2.00
kB /(mS/cm)

1.50
kA /(mS/cm)

1.50
1.00
kB =2.11 w(ash)+0.15 1.00
0.50
R 2=0.97
kA =2.03 w(ash)+0.14
0.50
0.00 R 2=0.97
0.00 0.20 0.40 0.60 0.80 1.00 1.20
0.00
w(ash)/% 0.00 0.20 0.40 0.60 0.80 1.00 1.20
w(ash)/%
Fig 2. Linear relationship between the kB and w(ash) in Slove-
nian honey (290 samples) Fig 4. Linear relationship between the kA and w(ash) in Slove-
nian honey

When comparing our model with the one proposed

by IHC, it becomes obvious that the method A for k de- The final regression model (y=ax+b), as presented in
termination of electrical conductivity should be used be- Fig. 4, between the mass fraction of ash and kA or kB
cause it is recommended in the harmonised methods (k=2.03 w(ash)+0.14) differs from the one proposed by
and was used when constructing the Italian model. the International Honey Commission (IHC): k=1.74
However, more time and glassware are needed for this w(ash)+0.14, where k is the electrical conductivity in
method. In order to save time, method B was used. mS/cm and ash is the ash mass fraction in g per 100 g
Sancho et al. (13) also used some simplifications for the (7). The slope differs greatly, but the section part of the
k determination but they also needed to use proper cor- relation formula does not. The relation was gathered by
rections when comparing their results to the results of analysing 290 samples of different pure types of Sloveni-
analyses performed according to IHC (7). Our modified an honey and their blends.
method for the determination of electrical conductivity As an illustration: if a sample of Slovenian honey
is faster than others and less glassware is needed. We at- has k=1.0 mS/cm, the ash mass fraction is 0.42 g per 100 g
tempted to assess differences between the two methods. as calculated by the Slovenian model, or 0.49 g per 100 g
Preliminary results of measurements on 31 honey sam- when using the model proposed by the IHC. The differ-
ples revealed that significant differences exist between ence of up to 17 % in ash mass fraction is obtained when
 U. KROPF et al.: Linear Regression Model for Slovenian Honey, Food Technol. Biotechnol. 46 (3) 335–340 (2008) 339

the different models are compared. Observed disparity different procedures. Therefore, a uniform standard in
is an outcome of a different type and a different geo- honey analysis provided by International Honey Com-
graphical origin of honey used in the research of Piazza mission should be used in as many laboratories as pos-
et al. (4) compared to our present work. sible around Europe. The exposed problem of the im-
Measuring the kA value is prescribed to take place portance of measuring the ash mass fraction in dry mat-
in 20 % (by mass per volume) water solution of honey, ter when establishing models of relationship between k
and is therefore always calibrated using the dry mass of and ash mass fraction should be discussed with leading
honey. But the mass fraction of ash is determined in the European researchers in the field of honey researches.
natural honey, not in its dry matter content. It would
thus be wise to make a comparison between k and the Acknowledgement
mass fraction of ash in the dry matter of honey. The mo- The authors would like to thank Mojca Ocepek for
del of linear regression for k and ash mass fraction in her technical assistance. We are also grateful to all bee-
dry matter (k=2.41 w(ash)+0.14) is presented in Fig. 5. keepers that provided honey samples.

2.50
References
2.00
1. G. Vorwohl, The relation between the electrical conducti-
kA /(mS/cm)

vity of honey and its botanical origin, Ann. Abeille, 7 (1964)

1.50
301–309 (in German).
2. G. Vorwohl, The measurement of the electrical conducti-
1.00
vity of honey and the use of the measuring values for ho-
kA =2.41 w(ash)+0.14 ney type determination and as a proof of falsifications with
0.50
R 2=0.97 honey from sugar fed bees, Z. Bienenforsch. 7 (1964) 37–47
0.00 (in German).
0.00 0.20 0.40 0.60 0.80 1.00 3. J. Louveaux, P. Pourtallier, G. Vorwohl, Methods of analy-
ses of honey. Conductivity, Bull. Apic. 16 (1973) 7 (in French).
w(ash)/% of dry matter
4. M.G. Piazza, M. Accorti, L. Persano Oddo, Electrical con-
Fig 5. Linear relationship between the kA and w(ash) of dry ductivity, ash, colour and specific rotatory power in Italian
matter for Slovenian honey unifloral honeys, Apicoltura, 7 (1991) 51–63.
5. M.T. Sancho, S. Muniategui, M.P. Sanchez, J.F. Huidobro, J.
Simal-Lozano, Study of ash in honey, Bee Science, 2 (1992)
This model has steeper slope and the same section part 147–153.
as the model without correction for water content, which 6. M.L. Felsner, C.B. Cano, R.E. Bruns, H.M. Watanabe, L.B.
was presented in Fig. 4, and is even less comparable to Almeida-Muradian, J.R. Matos, Characterization of mono-
the IHC model. The use of this model should be further floral honeys by ash contents through a hierarchical de-
discussed with other researchers in this field, especially sign, J. Food Compos. Anal. 17 (2004) 737–747.
since Sancho et al. (13) claimed that the effect of mois- 7. S. Bogdanov, P. Martin, C. Lüllmann, Harmonised methods
of the European Honey Commission, Apidologie, 28 (1997)
ture content on k determination is very small and there-
1–59.
fore can be neglected.
8. M. Accorti, M.G. Piazza, L. Persano Oddo, Electrical con-
ductivity and ash in honeys, Apicolt. Mod. 77 (1986) 165–167
(in Italian).
Conclusions
9. M.T. Sancho, S. Muniategui, M.P. Sánchez, J.F. Huidobro, J.
When comparing our model with the proposed one, Simal-Lozano, Relationships between electrical conducti-
it can be concluded that Slovenian honey obviously has vity and total and sulphated ash contents in Basque ho-
neys, Apidologie, 22 (1991) 487–494.
higher k values at the same ash mass fraction, which
10. T. Golob, U. Dober{ek, P. Kump, M. Ne~emer, Determina-
might be explained by its higher acid content. Therefore,
tion of trace and minor elements in Slovenian honey by
when calculating the ash mass fraction from k in the total reflection X-ray fluorescence spectroscopy, Food Chem.
samples of Slovenian honey, it is recommended that the 91 (2005) 593–600.
Slovenian formula be used for this relationship. The Slo- 11. M.T. Sancho, S. Muniategui, M.P. Sanchez, J.F. Huidobro, J.
venian model is valid in a range of k from 0.16 to 2.25 Simal-Lozano, Evaluating soluble and insoluble ash, alka-
mS/cm. Honey samples with k lower than 0.16 mS/cm linity of soluble and insoluble ash and total alkalinity of
are very rare in Slovenia, so this is not really a weak ash in honey using electrical conductivity measurements
point of this model. at 20 oC, Apidologie, 23 (1992) 291–297.
A faster and less complicated method B for kB deter- 12. G. Downey, K. Hussey, J.D. Kelly, T.F. Walshe, P.G. Martin,
Preliminary contribution to the characterisation of artisa-
mination can only be used if a proper correction factor
nal honey produced on the island of Ireland by palynolo-
is employed for recalculating the correct value. In Fig. 4, gical and physico-chemical data, Food Chem. 91 (2005)
the linear regression model of the kB and ash mass frac- 347–354.
tion in Slovenian honey are presented. The latter model 13. M.T. Sancho, S. Muniategui, J.F. Huidobro, J. Simal, Corre-
should be used for ash mass fraction calculation in our lation between the electrical conductivity of honey in hu-
laboratory, but we use the modified method for k deter- mid and dry matter, Apidologie, 22 (1991) 221–227.
mination. However, all simplifications of the honey so- 14. Determination of ash in honey, AOAC 920.181, Official
lution preparation may represent difficulties when com- Methods of Analysis of AOAC International, AOAC Inter-
paring the results among different laboratories that use national, Gaithersburg, MD, USA (1996).
340 U. KROPF et al.: Linear Regression Model for Slovenian Honey, Food Technol. Biotechnol. 46 (3) 335–340 (2008)

15. S. Bogdanov, Honey quality and international regulatory 18. A. Terrab, A.F. Recamales, D. Hernanz, F.J. Heredia, Cha-
standards: Review by the International Honey Commission, racterisation of Spanish thyme honeys by their physicoche-
Bee World, 80 (1999) 61–69. mical and mineral contents, Food Chem. 88 (2004) 537–542.
16. T. Golob, J. Bertoncelj, V. [krabanja, Sensory analysis of Slo- 19. J.L. Rodriguez-Otero, P. Paseiro, J. Simal, L. Terradillos, A.
vene honey, Proceedings of the 4th Croatian Congress of Food Cepeda, Silicon, phosphorus, sulphur, chlorine and ash
Technologists, Biotechnologists and Nutritionists – Central Eu-
contents of Spanish commercial honeys, Z. Lebensm. Un-
ropean Meeting, B. Tripalo (Ed.), Zagreb, Faculty of Food
ters. Forsch. A, 200 (1995) 233–234.
Technology and Biotechnology, University of Zagreb, Za-
greb, Croatia (2002) pp. 468–473. 20. V. Nanda, B.C. Sarkar, H.K. Sharma, A.S. Bawa, Physico-
17. @. Trstenjak Petrovi}, M.L. Mandi}, J. Grgi}, Z. Grgi}, Ash -chemical properties and estimation of mineral content in
and chromium levels of some types of honey, Z. Lebensm. honey produced from different plants in Northern India, J.
Unters. Forsch. A, 198 (1994) 36–39. Food Compos. Anal. 16 (2003) 613–619.