© Author(s) 2022. CC Atribution 4.0 LicenseGEOLOGIJA 65/2, 225-235, Ljubljana 2022
https://doi.org/10.5474/geologija.2022.013
Statistical approach to interpretation of geochemical data of 
stream sediment in Pleše mining area 
Statistični pristop k interpretaciji geokemičnih podatkov potočnega sedimenta na 
območju rudišča Pleše
Simona JARC
University of Ljubljana, Faculty of Natural Sciences and Engineering, Department of Geology, Aškerčeva c. 12,  
SI-1000 Ljubljana, Slovenia; e-mail: simona.jarc@ntf.uni-lj.si
Prejeto / Received 25. 10. 2022; Sprejeto / Accepted 7. 10. 2022; Objavljeno na spletu / Published online 21. 12. 2022
Key words: ANOVA, t-test, correlation, cluster analysis, XRF, mineralization
Ključne besede: ANOVA, t-test, korelacija, clustrska analiza, XRF, mineralizacija
Abstract
The Ba, Pb and Zn ore deposit Pleše near Ljubljana is one of the formerly productive mines. The stream sediments 
were sampled and analysed by XRF to establish the effect of grain size, mineralization, and downstream location 
of sampling sites on geochemical composition based on various statistical analyses. Statistical analyses of the 
geochemical data confirm the impact of mineralization. The parametric t-test, non-parametric Mann-Whitney 
test and cluster analysis showed only minor differences in the geochemical composition of the samples with 
different grain sizes (< 0.063 mm and 0.063-2 mm). The parametric and non-parametric correlation coefficients 
as well as cluster analysis indicate that the contents of Si, Al, K, Rb, and Fe are associated with weathered 
rock forming minerals such as micas, and clay minerals, whereas Nb and Zr are associated with minerals 
resistant to weathering. Ca and Mg are associated with carbonates. S, Ba, Sr, Pb, Zn, and Mn indicate local 
mineralization with sulphates and sulphides. The results of the t-test and analysis of variance, Mann-Whitney 
tests and Kruskal-Wallis ANOVA of the groups established by the cluster analysis confirm that the contents of 
Ba, Pb and Sr have a statistically significant influence on the classification of the cluster group - i.e., the influence 
of sediment mineralization. There are no differences in elemental contents in the sediment samples downstream. 
The statistical approach to evaluate the geochemical data has proven useful and provides a good basis for further 
interpretation. 
 
Izvleček
V rudišču Pleše v okolici Ljubljane so v preteklosti pridobivali Ba, Pb in podrejeno Zn. Vpliv velikosti 
zrn, mineralizacije in lokacije vzorčevanih točk dolvodno na geokemično sestavo sem določala v vzorcih 
potočnega sedimenta, katerega geokemična sestava je bila določena z metodo XRF. Statistična analiza podatkov 
geokemične sestave je potrdila mineralizacijo vzorcev potočnega sedimenta. Rezultati parametričnega t-testa, 
neparametičnega Mann-Whitney-jevega testa in clustrske analize so pokazali, da se geokemični sestavi vzorcev 
frakcij <0,063 mm in 0,063-2 mm skoraj ne razlikujeta. Tako parametrični kot neparametrični korelacijski 
koeficienti ter clustrska analiza so pokazali, da prvine Si, Al, K, Rb in Fe kažejo na prisotnost preperelih 
kamninotvornih mineralov (npr. sljud, glinenih mineralov), Nb in Zr na minerale odporne na preperevanje, Ca in 
Mg sta značilna za karbonate. S, Ba, Sr, Pb, Zn in Mn kažejo na lokalno mineralizacijo s sulfati in sulfidi. Razlike 
med skupinami, ugotovljenimi s clustrsko analizo, sem testirala s parametričnim t-testom, analizo variance in 
Mann-Whitney-jevim testom ter Kruskall-Wallis ANOVO. Na uvrstitev v skupine nabolj vplivajo vsebnosti Ba, 
Pb in Sr, kar potrjuje vpliv mineralizacije sedimenta. Razlik v geokemični sestavi vzorcev sedimentov dolvodno 
testi niso zaznali. Statistični pristop k interpretaciji geokemičnih podatkov predstavlja dobro osnovo za nadaljnjo 
interpretacijo. 
226 Primož MIKLAVC & †Bogomir CELARC
Introduction
Nowadays, statistical analyses are very com-
mon and useful in many fields of geology (Swan 
and Sandilands, 1995; Davis, 2002), especially 
when dealing with geochemical data (e.g., Al-
banese et al., 2007; Grunsky et al., 2009). A statis-
tical approach can also be useful in determining 
relationships between the geochemical compo-
sition of soils or stream sediments and mineral 
occurrences (e.g., Candeias et al., 2011; Carvalho 
et al., 2014; Levitan et al., 2015). There are some 
former mines and small ore deposits of Pb, Zn, 
Hg and Fe in the vicinity of Ljubljana (Dervarič 
et al., 2005). The Pleše Ba, Pb, and Zn ore deposit 
is one of the formerly productive mines (Mlakar, 
2003). Barite, galena, and some sphalerite were 
mined here for at least 250 years. Between 1729 
and 1963, when the mine was opened, more than 
100,000 tons of barite, about 10,000 tons of Pb, 
and some Zn were mined (Žebre, 1955; Fabjančič, 
1966; Mlakar, 2003). Today, water seeps and leaks 
from the formerly productive mine and trans-
ports mineralized sediments that may affect the 
local environment. Stream sediment chemistry 
is one of the indicators of local geology and also 
of mining activity (e.g., Hudson-Edwards et al., 
1996; Ettler et al., 2006; Teršič et al., 2018). It can 
also be used as an indicator of potential contam-
ination from mining materials (Baptista-Salazar 
et al., 2017; Potra et al., 2017; Gosar et al., 2020; 
Žibret & Čeplak, 2021; Miler et al., 2022).
Sediments from the stream of the abandoned 
mine tunnel in the Pleše area were sampled to 
evaluate their geochemical properties and the in-
fluence of grain size (0.063-2 mm and < 0.063 mm) 
based on statistical analyses. In addition, statis-
tical tests were performed to evaluate the impact 
of the sampling site on sediment geochemical 
composition.
Materials and methods
The oldest rocks of the area are the Carbon-
iferous clastites (Buser, 1974; Mlakar, 2003). The 
Permian rocks of the Val Gardena Formation 
are followed by Lower Triassic dolomites inter-
bedded with fine clastites, red claystones, oolitic 
limestone and dolomite lenses. This is followed 
by Anisian and Cordevolian dolomites, and fi-
nally the upper Triassic Main dolomite. Stream 
and bog sediments with scree are the youngest 
rocks in the area (Buser et al., 1969; Buser, 1974; 
Mlakar 2003). The area has a complex tectonic 
history (Buser, 1974; Premru 1974, 1980; Mlakar, 
1987; Placer, 1998; Dozet, 1999; Mlakar, 2003). The 
Upper Triassic dolomite is overthrusted by Pale-
ozoic beds. In addition, several faults of different 
systems (e.g., cross-alpidic, dinaric, cross-dinar-
ic) are observed. Mineralization with barite, ga-
lena and sphalerite occurs in both Paleozoic and 
Triassic beds (Buser, 1974; Mlakar 2003).
Stream sediment samples were collected at 10 
sampling sites at a total distance of 250 m from 
the abandoned mining tunnel (Fig. 1). The upper 
30 cm of the stream sediment was sampled. In the 
laboratory, all samples were dried at 40 °C. Sam-
ple size was reduced by quartering. Samples were 
sieved through sieves with a 2-mm and a 0.063-
mm openings. A total of 15 samples were analysed 
- 5 samples of fraction 0.063-2 mm (samples des-
ignated PO1-2, PO3-2, PO5-2, PO7-2, PO9-2) and 
10 samples with grain size of < 0.063 mm (samples 
designated PO1 to PO10). Geochemical composi-
tion was determined using a Thermo Fisher Niton 
XL3t GOLDD portable X-ray fluorescence (XRF) 
analyzer at the Department of Geology, Faculty of 
Natural Sciences and Engineering, University of 
Ljubljana. The accuracy was checked with sever-
al analyses of standard materials and found to be 
good for elements in question. The only exception 
is Ba, whose values were not standardised and 
therefore should be used with caution and only for 
relative comparison of the analysed samples. Re-
lative percentage difference (%RPD) of duplicate 
measurements of sample PO1 showed very good 
precision (< 10 % error) for most elements, with the 
exception of S (< 12 % error). For further interpre-
tation, the contents of Si, Al, Fe, Mg, Ca, K, S, Mn, 
Ba, Nb, Pb, Rb, Sr, Zn, and Zr were manipulated. 
The contents of the other elements were below the 
detection limit in the majority of the samples or 
were not of our interest. Statistical analysis was 
performed using Tibco Statistica software (2017). 
Normality of data distribution was checked by 
comparison of medians, arithmetic and geometric 
means, by kurtosis and skewness, by visual in-
spection of histograms, by Kolmogorov-Smirnov 
and Lilliefors test, and by Shapiro-Wilk test. Be-
cause some of the variables aren’t normally dis-
tributed and because of the small number of sam-
ples and for comparison, we applied parametric 
and non-parametric statistics (Swan and Sandi-
lands, 1995; Davis, 2002). For comparision para-
metric t-test and non-parametric Mann-Whitney 
test were performed and correlations (Pearsoǹ  s 
product-moment and Spearman rank order cor-
relation coefficients) were calculated. Cluster 
analyses of the variables and observations were 
performed using Ward’s linkage rule method and 
the 1-Pearson correlation coefficient and Euclid-
ean distance as a distance measure. 
227Depositional environment of the Middle Triassic Strelovec Formation on Mt. Raduha, Kamnik-Savinja Alps, northern Slovenia
Results and Discussion
Ba-Pb-Zn mineralization is clearly evident in 
the geochemical compositions of the stream sedi-
ment samples (Table 1). In some samples, the con-
tents of Ba (the values reported should be used 
with caution and for relative comparison only), 
Pb, and Zn exceed the intervention values of 625, 
530, and 720 mg/kg, respectively, in soils (Ur. list 
RS 68/96) and in soils and sediments (VROM, 
2000). The contents of Ba and Pb above interven-
tion values were also established by others (Go-
sar et al., 2014; Miler et al., 2022). For compari-
sion, median values for Ba, Pb and Zn in stream 
sediments analysed by XRF in Europe are 370, 21 
and 71 mg/kg (Salminen et al., 2005).
The presence of silicate and carbonate mine-
rals in sediment samples is also evident (Table 1). 
The variations in geochemical composition is  the 
result of complex geological composition of the 
area (Buser, 1974; Mlakar, 2003).
The effect of grain size on the distribution 
of elements was examined using the parametric 
t-test and the non-parametric Mann-Whitney 
test. The parametric t-test revealed statistically 
significant differences (95 % probability) in the 
studied fractions (0.063-2 mm and < 0.063 mm) 
with respect to the content of Mg (t = -2.53, p = 
0.025), Nb (t = 2.26, p = 0.041), and Zr (t = 2.35, 
p = 0.036). The results of the Mann-Whitney test 
were very similar, with statistically significant 
differences at the 95 % probability level for the 
contents of Nb (Z = 2.02, p = 0.043) and Zr (Z = 
2.88, p = 0.004) in the studied grain size fractions. 
The results are also confirmed by box and whisk-
er plots (Fig. 2) - the geochemical composition 
is very similar in both fractions. Some samples 
are more mineralized than others, with no clear 
trend as a function of grain size. Minor differ-
ences are observed only in Mg content, which is 
almost twice as high in the coarse-grained frac-
tion, with a mean of 9.01 % and a median of 9.73 % 
(in the < 0.063 mm samples, the mean is 5.44 % 
and the median is 5.98 %). The Mg content can 
be attributed to the presence of dolomite. Niobi-
um and Zr are more abundant in the fine-grained 
fraction. For Nb, the mean is 16 mg/kg and the 
median is 15 mg/kg in the < 0.063 mm fraction, 
while it is lower in the coarser fraction at 9 mg/
kg and 7 mg/kg, respectively. The difference in 
Zr content is even greater: mean and median are 
251 mg/kg and 150 mg/kg in the finer fraction, 
while 42 mg/kg (mean) and 41 mg/kg (median) in 
the coarser fraction. Niobium and Zr are bound 
in weathering resistant minerals. 
The small number of samples and the (non-)
normality of the distribution clearly affect the 
differences between mean and median values 
(Swan & Sandilands, 1995). In the present case, 
the number of samples (10 vs. 5 samples) also af-
fects the range - the ranges of the fine-grained 
samples (< 0.063 mm) are generally larger than 
the ranges of the 0.063-2 mm samples for most 
elements. The exceptions with wide ranges for 
S, Ba, Pb, Sr, and Zn in the coarser fraction are 
probably due to the mineralization of the sedi-
ments. 
Fig. 1. Map of the area 
with sampling locations. 
228 Simona JARC
Fig. 2. Medians, interquartile ranges (boxes) with minimum and maximum values (whiskers) of measured elements in finer 
(<0.063 mm) and coarser (0.063-2 mm) fractions.
all data to calculate the parametric correlation 
coefficient. If one or more variable values are 
missing, the entire observation is excluded from 
the analysis. As Mn content in sample PO9-2 was 
not detected by XRF, the entire observation was 
eliminated from the analysis. Therefore, when 
The relations between elements have been 
determined by calculation of correlation coef-
ficients and by cluster analysis. Care should be 
taken when comparing parametric and non-par-
ametric correlation coefficients calculated using 
Tibco Statistica software. The software requires 
229Statistical approach to interpretation of geochemnical data of stream sediment in Pleše mining area
S
am
p
le
G
ra
in
 s
iz
es
 
(m
m
)
S
i 
(%
)
A
l 
(%
)
F
e 
(%
)
M
g 
(%
)
C
a 
(%
)
K
 (
%
)
S
 (
%
)
M
n
 (
%
)
B
a 
(%
)
N
b
 (
m
g/
k
g)
P
b
 (
m
g/
k
g)
R
b
 (
m
g/
k
g)
S
r 
(m
g/
k
g)
Z
n
 (
m
g/
k
g)
Z
r 
(m
g/
k
g)
P
O
1
<0
.0
63
26
.9
89
7.
75
0
2.
95
0
1.
47
0
2.
68
0
1.
98
8
0.
18
8
0.
12
8
1.
18
8
27
24
5
10
5
30
2
30
8
46
0
P
O
2
<0
.0
63
22
.0
58
4.
88
2
2.
12
4
2.
48
2
5.
47
5
1.
32
3
0.
38
9
0.
10
6
1.
65
0
22
37
6
69
40
9
31
9
55
4
P
O
3 
<0
.0
63
24
.0
15
5.
29
6
2.
00
9
2.
49
3
6.
10
4
1.
45
0
0.
47
5
0.
15
9
2.
06
1
24
43
5
71
45
9
44
5
57
4
P
O
4
<0
.0
63
10
.2
40
2.
92
0
1.
53
1
5.
98
9
15
.4
50
0.
76
5
0.
48
2
0.
18
8
2.
27
5
15
14
54
44
52
5
66
4
16
1
P
O
5 
<0
.0
63
7.
66
5
2.
56
1
1.
26
9
5.
96
7
19
.3
52
0.
69
7
0.
46
5
0.
48
9
2.
83
8
14
15
20
37
61
5
83
9
13
8
P
O
6
<0
.0
63
9.
20
1
3.
06
0
1.
34
1
6.
84
3
15
.9
70
0.
70
0
0.
53
1
0.
16
6
2.
56
4
15
95
4
34
58
4
34
2
13
8
P
O
7 
<0
.0
63
8.
79
0
2.
72
0
1.
12
4
7.
50
3
16
.3
94
0.
59
5
0.
41
0
0.
10
8
1.
92
8
12
68
8
32
42
5
28
1
12
6
P
O
8
<0
.0
63
5.
59
1
1.
32
4
0.
96
7
4.
65
5
15
.6
39
0.
45
6
0.
40
0
0.
10
4
1.
90
8
13
56
7
26
48
2
19
9
12
4
P
O
9 
<0
.0
63
5.
99
7
1.
76
7
0.
80
3
9.
05
1
18
.7
15
0.
36
3
0.
21
3
0.
03
7
0.
70
3
9
12
48
21
20
6
21
5
76
P
O
10
<0
.0
63
8.
53
1
2.
29
9
1.
04
5
7.
90
7
16
.7
42
0.
52
1
0.
35
6
0.
07
1
1.
72
7
12
11
93
30
43
6
25
0
16
2
P
O
1-
2
0.
06
3-
2
16
.9
70
3.
64
3
2.
33
0
5.
35
6
9.
56
5
0.
98
4
0.
05
6
0.
11
2
0.
11
7
9
17
7
41
77
15
0
83
P
O
3-
2
0.
06
3-
2
11
.0
80
2.
18
3
1.
53
5
7.
43
8
14
.2
74
0.
63
3
0.
30
0
0.
07
4
0.
63
4
7
37
1
24
18
6
76
9
53
P
O
5-
2
0.
06
3-
2
5.
91
0
1.
58
0
1.
40
1
9.
72
6
16
.2
84
0.
41
6
2.
10
6
0.
43
2
7.
23
9
18
23
69
28
13
50
54
4
41
P
O
7-
2
0.
06
3-
2
2.
64
9
0.
85
3
0.
54
9
11
.4
82
19
.9
40
0.
23
5
0.
45
3
0.
07
1
1.
28
5
7
38
7
12
34
7
12
2
19
P
O
9-
2
0.
06
3-
2
1.
55
3
0.
51
0
0.
36
1
11
.0
44
21
.6
51
0.
11
6
0.
10
7
0.
05
5
4
19
7
5
10
2
40
11
T
ab
le
 1
. R
es
u
lt
s 
of
 g
eo
ch
em
ic
a
l 
a
n
a
ly
si
s.
comparing the two correlation coefficients, some 
data are “lost” - regardless of whether we omit 
the entire observation or the variable with the 
missing data. Consequently, the results may dif-
fer to some extent and may also be biased.
First, the coefficients were calculated without 
the sample PO9-2 data. For most elements, there 
are almost no differences in the statistically sig-
nificant correlation coefficients calculated with 
the parametric Pearsoǹ s product-moment corre-
lation coefficient (Table 2) or with non-paramet-
ric Spearman rank order correlation coefficient 
(Table 3). Exceptions are Si-Nb, Fe-Zr, and Mn-
Pb, for which parametric correlations were deter-
mined to be statistically significant, and Si-Pb, 
Mn-Nb, Mn-Rb, and Ba-Zn with statistically sig-
nificant non-parametric correlation coefficients. 
The results look slightly different if we omit 
the Mn values (Tables 4 and 5) instead of PO9-2 
observation. In this case, the differences are in 
S-Zn, Ba-Nb, Ba-Zn, Nb-Sr, and Sr-Zn correla-
tions, where non-parametric correlation coeffi-
cients are statistically significant, and Pearsoǹ s 
product-moment correlation coefficients areǹ t. 
The Pearsoǹ s product-moment correlation coef-
ficient is based on the agreement and direction 
of the linear relationship between two variables, 
while the non-parametric correlation is based on 
the ranking of the data values rather than the 
values themselves (Swan & Sandilands, 1995). 
Therefore, scatter plots with a distinct trend line 
may have a low non-parametric correlation even 
though the parametric correlation is statistically 
significant because the positive and negative de-
viations from the line almost cancel each other 
out. The non-parametric correlation is also less 
sensitive to extreme values (Swan & Sandilands, 
1995). On the other hand, the calculation of Pear-
soǹ s product-moment correlation coefficient re-
quires a larger number of samples and normality 
of the distribution. Therefore in presented case, 
the results of non-parametric correlations are 
more trustworthy, while parametric Pearsoǹ s 
product-moment correlation coefficient should 
be used with caution.
The results of cluster analysis of the var-
iables show two main groups (Fig. 3). The first 
group consists of Si, Al, K, Rb, Fe, Nb, and Zr, 
while the second group can be divided into two 
subgroups, with Ca and Mg in one and Zn, Pb, 
Mn, Sr, Ba, and S in the other subgroup. Cluster 
analysis confirms the results of correlation coef-
ficients. Namely, members of the groups have in 
general higher correlation coefficients. The con-
tents of Si, Al, K, Rb, and Fe can be attributed 
230 Simona JARC
S
i 
(%
)
S
i 
(%
)
A
l 
(%
)
0.
93
A
l 
(%
)
F
e 
(%
)
0.
89
0.
81
F
e 
(%
)
M
g 
(%
)
-0
.7
2
-0
.7
5
-0
.7
0
M
g 
(%
)
C
a 
(%
)
-0
.8
5
-0
.7
4
-0
.8
9
0.
80
C
a 
(%
)
K
 (
%
)
0.
93
0.
96
0.
88
-0
.8
5
-0
.8
2
K
 (
%
)
S
 (
%
)
-0
.3
0
-0
.1
5
-0
.2
0
0.
24
0.
27
-0
.1
3
S
 (
%
)
M
n
 (
%
)
0.
25
0.
41
0.
45
-0
.3
2
-0
.2
1
0.
47
0.
60
M
n
 (
%
)
B
a 
(%
)
-0
.2
6
-0
.0
4
-0
.1
6
0.
09
0.
26
-0
.0
4
0.
91
0.
73
B
a 
(%
)
N
b
 (
m
g
/k
g
)
0.
48
0.
63
0.
55
-0
.6
2
-0
.5
5
0.
64
0.
35
0.
64
0.
50
N
b
 (
m
g
/k
g
)
P
b
 (
m
g
/k
g
)
-0
.5
6
-0
.3
9
-0
.5
0
0.
48
0.
61
-0
.4
5
0.
65
0.
37
0.
76
0.
11
P
b
 (
m
g
/k
g
)
R
b
 (
m
g
/k
g
)
0.
82
0.
93
0.
79
-0
.8
1
-0
.7
1
0.
94
0.
01
0.
59
0.
17
0.
79
-0
.2
3
R
b
 (
m
g
/k
g
)
S
r 
(m
g
/k
g
)
-0
.3
3
-0
.1
3
-0
.2
0
0.
07
0.
27
-0
.0
9
0.
87
0.
68
0.
98
0.
49
0.
77
0.
11
S
r 
(m
g
/k
g
)
Z
n
 (
m
g
/k
g
)
0.
29
0.
24
0.
37
-0
.1
6
-0
.1
7
0.
34
0.
48
0.
67
0.
54
0.
38
0.
39
0.
32
0.
49
Z
n
 (
m
g
/k
g
)
Z
r 
(m
g
/k
g
)
0.
68
0.
80
0.
48
-0
.7
1
-0
.5
4
0.
77
0.
02
0.
29
0.
20
0.
73
-0
.1
0
0.
83
0.
17
0.
25
T
ab
le
 2
. P
ea
rs
on
 s
 p
ro
d
u
ct
-m
om
en
t 
co
rr
el
at
io
n
 c
o
ef
fi
ci
en
ts
 (
re
d
 m
a
rk
ed
 c
or
re
la
ti
on
s 
a
re
 s
ig
n
ifi
ca
n
t 
at
 p
 <
0.
05
).
T
ab
le
 3
. N
on
-p
a
ra
m
et
ri
c 
S
p
ea
rm
a
n
 r
a
n
k
 o
rd
er
 c
or
re
la
ti
on
 c
o
ef
fi
ci
en
ts
 (
re
d
 m
a
rk
ed
 c
or
re
la
ti
on
s 
a
re
 s
ig
n
ifi
ca
n
t 
at
 p
 <
0.
05
).
S
i 
(%
)
S
i 
(%
)
A
l 
(%
)
0.
96
A
l 
(%
)
F
e 
(%
)
0.
92
0.
91
F
e 
(%
)
M
g 
(%
)
-0
.8
7
-0
.8
4
-0
.8
0
M
g 
(%
)
C
a 
(%
)
-0
.9
7
-0
.9
1
-0
.9
2
0.
86
C
a 
(%
)
K
 (
%
)
0.
97
0.
99
0.
93
-0
.8
8
-0
.9
4
K
 (
%
)
S
 (
%
)
-0
.3
0
-0
.2
9
-0
.1
6
0.
36
0.
22
-0
.2
8
S
 (
%
)
M
n
 (
%
)
-0
.1
5
-0
.0
7
0.
03
0.
04
0.
20
-0
.0
5
0.
64
M
n
 (
%
)
B
a 
(%
)
-0
.2
7
-0
.2
2
-0
.1
4
0.
24
0.
22
-0
.2
2
0.
97
0.
76
B
a 
(%
)
N
b
 (
m
g
/k
g
)
0.
77
0.
82
0.
71
-0
.7
6
-0
.7
6
0.
83
0.
17
0.
25
0.
28
N
b
 (
m
g
/k
g
)
P
b
 (
m
g
/k
g
)
-0
.5
2
-0
.4
3
-0
.3
8
0.
46
0.
54
-0
.4
6
0.
74
0.
69
0.
80
-0
.0
4
P
b
 (
m
g
/k
g
)
R
b
 (
m
g
/k
g
)
0.
93
0.
98
0.
88
-0
.8
7
-0
.9
1
0.
98
-0
.1
9
0.
01
-0
.1
1
0.
90
-0
.3
5
R
b
 (
m
g
/k
g
)
S
r 
(m
g
/k
g
)
-0
.2
9
-0
.2
3
-0
.1
7
0.
24
0.
24
-0
.2
3
0.
95
0.
75
1.
00
0.
29
0.
80
-0
.1
1
S
r 
(m
g
/k
g
)
Z
n
 (
m
g
/k
g
)
-0
.0
2
-0
.0
1
0.
08
-0
.0
6
0.
12
0.
03
0.
28
0.
64
0.
34
0.
10
0.
43
0.
02
0.
34
Z
n
 (
m
g
/k
g
)
Z
r 
(m
g
/k
g
)
0.
87
0.
84
0.
67
-0
.8
4
-0
.8
4
0.
86
-0
.2
0
-0
.1
0
-0
.1
2
0.
86
-0
.3
8
0.
87
-0
.1
1
-0
.0
3
231Statistical approach to interpretation of geochemnical data of stream sediment in Pleše mining area
T
ab
le
 5
. N
on
-p
a
ra
m
et
ri
c 
S
p
ea
rm
a
n
 r
a
n
k
 o
rd
er
 c
or
re
la
ti
on
 c
o
ef
fi
ci
en
ts
 (
w
it
h
ou
t 
M
n
; r
ed
 m
a
rk
ed
 c
or
re
la
ti
on
s 
a
re
 s
ig
n
ifi
ca
n
t 
at
 p
 <
0.
05
).
S
i 
(%
)
S
i 
(%
)
A
l 
(%
)
0.
96
A
l 
(%
)
F
e 
(%
)
0.
93
0.
92
F
e 
(%
)
M
g 
(%
)
-0
.8
8
-0
.8
6
-0
.8
3
M
g 
(%
)
C
a 
(%
)
-0
.9
8
-0
.9
2
-0
.9
3
0.
88
C
a 
(%
)
K
 (
%
)
0.
98
0.
99
0.
93
-0
.9
0
-0
.9
5
K
 (
%
)
S
 (
%
)
-0
.2
0
-0
.1
9
-0
.0
6
0.
24
0.
13
-0
.1
9
S
 (
%
)
B
a 
(%
)
-0
.1
4
-0
.0
9
0.
01
0.
09
0.
09
-0
.0
9
0.
96
B
a 
(%
)
N
b
 (
m
g
/k
g
)
0.
80
0.
85
0.
76
-0
.8
0
-0
.8
0
0.
85
0.
24
0.
36
N
b
 (
m
g
/k
g
)
P
b
 (
m
g
/k
g
)
-0
.3
8
-0
.2
9
-0
.2
2
0.
29
0.
39
-0
.3
2
0.
76
0.
81
0.
08
P
b
 (
m
g
/k
g
)
R
b
 (
m
g
/k
g
)
0.
94
0.
98
0.
90
-0
.8
9
-0
.9
2
0.
98
-0
.1
0
0.
01
0.
91
-0
.2
1
R
b
 (
m
g
/k
g
)
S
r 
(m
g
/k
g
)
-0
.1
5
-0
.1
0
-0
.0
2
0.
08
0.
10
-0
.1
0
0.
95
1.
00
0.
37
0.
81
0.
01
S
r 
(m
g
/k
g
)
Z
n
 (
m
g
/k
g
)
0.
11
0.
12
0.
23
-0
.2
0
-0
.0
3
0.
15
0.
33
0.
42
0.
24
0.
49
0.
16
0.
41
Z
n
 (
m
g
/k
g
)
Z
r 
(m
g
/k
g
)
0.
88
0.
85
0.
69
-0
.8
5
-0
.8
5
0.
86
-0
.1
4
-0
.0
4
0.
86
-0
.2
9
0.
88
-0
.0
3
0.
07
S
i 
(%
)
S
i 
(%
)
A
l 
(%
)
0.
94
A
l 
(%
)
F
e 
(%
)
0.
91
0.
85
F
e 
(%
)
M
g 
(%
)
-0
.7
7
-0
.7
9
-0
.7
5
M
g 
(%
)
C
a 
(%
)
-0
.8
8
-0
.7
9
-0
.9
1
0.
83
C
a 
(%
)
K
 (
%
)
0.
95
0.
96
0.
90
-0
.8
8
-0
.8
6
K
 (
%
)
S
 (
%
)
-0
.1
0
0.
03
-0
.0
3
0.
08
0.
07
0.
04
S
 (
%
)
B
a 
(%
)
-0
.0
2
0.
15
0.
06
-0
.1
0
0.
02
0.
16
0.
92
B
a 
(%
)
N
b
 (
m
g
/k
g
)
0.
58
0.
70
0.
63
-0
.6
8
-0
.6
4
0.
71
0.
46
0.
60
N
b
 (
m
g
/k
g
)
P
b
 (
m
g
/k
g
)
-0
.3
0
-0
.1
7
-0
.2
6
0.
25
0.
35
-0
.2
1
0.
71
0.
80
0.
26
P
b
 (
m
g
/k
g
)
R
b
 (
m
g
/k
g
)
0.
85
0.
95
0.
83
-0
.8
4
-0
.7
6
0.
95
0.
16
0.
33
0.
83
-0
.0
4
R
b
 (
m
g
/k
g
)
S
r 
(m
g
/k
g
)
-0
.1
2
0.
05
-0
.0
3
-0
.0
8
0.
07
0.
08
0.
90
0.
98
0.
57
0.
81
0.
24
S
r 
(m
g
/k
g
)
Z
n
 (
m
g
/k
g
)
0.
43
0.
38
0.
49
-0
.3
1
-0
.3
3
0.
46
0.
57
0.
63
0.
50
0.
50
0.
45
0.
58
Z
n
 (
m
g
/k
g
)
Z
r 
(m
g
/k
g
)
0.
74
0.
84
0.
58
-0
.7
6
-0
.6
3
0.
81
0.
19
0.
35
0.
78
0.
09
0.
86
0.
31
0.
39
T
ab
le
 4
. P
ea
rs
on
`s
 p
ro
d
u
ct
-m
om
en
t 
co
rr
el
at
io
n
 c
o
ef
fi
ci
en
ts
 (
w
it
h
ou
t 
M
n
; r
ed
 m
a
rk
ed
 c
or
re
la
ti
on
s 
a
re
 s
ig
n
ifi
ca
n
t 
at
 p
 <
0.
05
).
232 Simona JARC
to secondary minerals (e.g., clay minerals, micas) 
or oxides (hematite; Mlakar 2003), Nb and Zr to 
weathering-resistant minerals, and Ca and Mg to 
carbonates. Ba, Pb, Zn and S indicate local min-
eralization with sulphates and sulphides, while 
Sr might be attributed to trace elements in barite 
(Mlakar, 2003) and Mn in galenite and sphalerite 
(Drovenik et al., 1980) or to secondary minerals 
formed with ore mineral weathering (Miler et al., 
2022). 
When clustering the observations without 
sample PO9-2 (using Ward’s method as a linkage 
rule and Euclidean distance measurement), three 
groups are distinguished (Fig. 4): the first group 
with samples PO1, PO1-2, PO2, PO3, PO3-2, PO7-
2; the second group with samples PO6, PO7, PO8, 
PO9, PO10; and the third group with samples 
PO4, PO5, and PO5-2. We checked the differences 
between the groups obtained by cluster analysis 
using analysis of variance and Kruskal-Wallis 
Fig. 3. Cluster analysis of 
measured elements (calcu-
lation is based on Ward's 
method as a linkage rule 
and 1-Pearson r as a distan-
ce measurement).
Fig. 4. Cluster analysis of 
investigated samples witho-
ut PO9-2 (calculation is ba-
sed on Ward's method as a 
linkage rule and Euclidean 
distances as a distance 
measurement).
233Statistical approach to interpretation of geochemnical data of stream sediment in Pleše mining area
ANOVA test. The results show that the content 
of Pb, Ba, Mn, and Sr (i.e., the mineralization of 
the sediment) has the greatest influence on the 
classification into these three individual groups. 
If we exclude the Mn content from the analysis 
(it was not detectable in one sample) and manip-
ulate all 15 sediment samples, the result of the 
cluster analysis looks somewhat different (Fig. 
5). Only two groups can be distinguished: PO1, 
PO2, PO3, PO7, PO8, PO7-2, PO3-2, PO1-2 and 
PO9-2 form one group, PO4, PO5, PO6, PO9, 
PO10 and PO05-2 form the second. According 
to the results of the parametric t-test for inde-
pendent groups, the contents of Pb and Sr have a 
significant effect on the grouping. Due to the re-
sults of the non-parametric Mann-Whitney test, 
the contents of Pb, Ba and Sr have a statistically 
significant influence on the grouping of observa-
tions. The statistical analysis clearly shows the 
different contents of ore minerals of the samples 
and its effect on the clustering of observations. 
However, the grain size has no influence on the 
geochemical composition of the studied samples 
(< 0.063 mm and 0.063-2 mm; Figs. 4 and 5) as 
geochemical composition is practically the same 
in both fractions. Cluster analysis of the samples 
also shows no significant differences in the geo-
chemical composition regarding the downstream 
location of sediment samples.
Conclusions
Mineralization with barite, galena and 
sphalerite in the Pleše area is clearly demon-
strated in the geochemical composition of stream 
sediments. The contents of Ba (the values given 
should be used with caution and only for relative 
comparison), Pb and Zn in some samples exceed 
the intervention values of 625, 530 and 720 mg/
kg, respectively (Ur. list RS 68/96; VROM, 2000). 
The parametric t-test and the non-parametric 
Mann-Whitney test showed only minor differ-
ences in geochemical composition between sam-
ples with different grain sizes, whereas cluster 
analysis shows no differences. The Mg content 
is statistically significantly higher in samples of 
0.063-2 mm, and the contents of Nb and Zr are 
higher in < 0.063-mm samples. The number of ob-
servations (sediment samples) influence the rang-
es of the variables as the ranges of fine-grained 
samples are generally larger than ranges of 
coarse-grained samples. The diverse composition 
is the result of the complicated geological com-
position of the area (Buser, 1974; Mlakar, 2003). 
Ba, Sr, Pb, Zn, Mn and S indicate local mineral-
ization with sulphide and sulphate minerals and 
their weathering products.
Pearsoǹ s product-moment correlation coeffi-
cients, non-parametric Spearman rank order cor-
relation coefficients and cluster analysis indicate 
that Si, Al, K, Rb, and Fe contents are associated 
with weathered rock forming minerals such as 
Fig. 5. Cluster analysis of 
investigated samples witho-
ut Mn content in all samples 
(calculation is based on 
Ward' method as a linkage 
rule and Euclidean distances 
as a distance measurement).
234 Simona JARC
mica, and clay minerals, whereas Nb and Zr are 
associated with weathering-resistant minerals 
(oxides, silicates). Ca and Mg are characteristics 
of carbonates and S, Ba, Sr, Pb, Zn and Mn of lo-
cal mineralization with sulphides and sulphates. 
The results of the parametric t-test, analysis of 
variance, and non-parametric Mann-Whitney 
and Kruskal-Wallis ANOVA tests of the clustered 
sediment samples confirm that the contents of 
Ba, Pb and Sr have a statistically significant in-
fluence on the clustering of the sediment samples, 
i.e., the statistical analyses confirm the influence 
of sediment mineralization. Cluster analysis of 
the observations also shows no significant differ-
ences in the geochemical compositions regarding 
the downstream sampling position.
Although the number of sediment samples 
was relatively small, the combination of various 
statistical tests and analyses shows results that 
have a geologic basis and significance. The sta-
tistical approach to evaluating the geochemical 
data has proven useful and provides a good basis 
for assigning elemental data to parent rocks and 
for identifying potentially contaminated areas.
Acknowledgments
The author gratefully acknowledges financial 
support from the state budget through the Slo-
venian Research Agency (Programme No. P1-
0195). Special thanks to Ema Hrovatin, Primož 
Miklavc, Miran Udovč and Gašper M. Volk for 
preparation of the samples and technical support, 
and to Prof. Dr. Nina Zupančič for her comments 
and suggestions.
References
Albanese, S., De Vivo, B., Lima, A. & Cicchella, 
D. 2007: Geochemical background and base-
line values of toxic elements in stream sedi-
ments of Campania region (Italy). J Geochem 
Explor, 93: 21–34. https://doi.org/10.1016/j.
gexplo.2006.07.006
Baptista-Salazar, C., Richard, J., Horf, M., Rejc, 
M., Gosar, M. & Biester, H. 2017: Grain-size 
dependence of mercury speciation in river 
suspended matter, sediments and soils in 
a mercury mining area at varying hydro-
logical conditions. Appl Geochemistry, 
81: 132–142. https://doi.org/10.1016/j.
apgeochem.2017.04.006
Buser, S. 1974: Tolmač lista Ribnica, Osnovna ge-
ološka karta SFRJ 1:100.000. Zvezni geološki 
zavod, Beograd: 60 p.
Buser, S., Dozet, S., Cajhen, J., Ferjančič, L., 
Grad, K. et al. 1969: Osnovna geološka karta 
SFRJ 1: 100.000, list Ribnica. Zvezni geološki 
zavod, Beograd.
Candeias, C., Ferreira da Silva, E., Salgueiro, 
A.R., Pereira, H.G., Reis, A.P., Patinha, C., 
Matos, J.X. & Ávila, P.H. 2011: The use of 
multivariate statistical analysis of geoche-
mical data for assessing the spatial distri-
bution of soil contamination by potentially 
toxic elements in the Aljustrel mining area 
(Iberian Pyrite Belt, Portugal). Environ Earth 
Sci, 62: 1461–1479. https://doi.org/10.1007/
s12665-010-0631-2
Carvalho, P.C.S., Neiva, A.M.R., Silva, M.M.V.G. 
& Ferreira da Silva, E.A. 2014: Geochemical 
comparison of waters and stream sediments 
close to abandoned Sb-Au and As-Au mi-
ning areas, northern Portugal. Chemie der 
Erde, 74/2: 267–283. https://doi.org/10.1016/j.
chemer.2013.08.003
Davis, J.C. 2002: Statistics and data analysis in 
geology, 3rd Ed. John Wiley & Sons, New 
York: 638 p.
Dervarič, E., Herlec, U., Likar, J. & Bajželj, U. 
2005: Rudniki in premogovniki v Sloveniji. 
Argos, Nazarje: 175 p.
Dozet, S. 1999: Barite-bearing Pleše Formation, 
Central Slovenia Camparison of barite-bea-
ring beds and barite occurrences in the Outer 
Dinarides area (Pleška baritonosna forma-
cija, osrednja Slovenija: Primerjava barito-
nosnih plasti in baritnih pojavov na območju 
Zunanjih Dinaridov). Geologija, 42/1: 41–68. 
https://doi.org/10.5474/geologija.1999.004
Drovenik, M., Pleničar, M. & Drovenik, F. 1980: 
Nastanek rudišč v SR Sloveniji = The origin 
of Slovenian ore deposits. Geologija, 23/1: 
1–137. 
Ettler, V., Mihaljevič, M., Šebek, O., Molek, M., 
Grygar, T. & Zeman, J. 2006: Geochemical 
and Pb isotopic evidence for sources and 
dispersal of metal contamination in stream 
sediments from the mining and smelting di-
strict of Příbram, Czech Republic. Environ 
Pollut, 142/3: 409–17. https://doi.org/10.1016/j.
envpol.2005.10.024
Fabjančič, M. 1966: About barite occurrences in 
Slovenia (O baritu na Slovenskem). Geologija, 
9/1: 505–526. https://www.geologija-revija.si/
index.php/geologija/article/view/234
Gosar, M., Šajn, R., Miler, M., Burger, A., & Bavec, 
Š. 2020: Overview of existing information on 
important closed (or in closing phase) and 
abandoned mining waste sites and related 
235Statistical approach to interpretation of geochemnical data of stream sediment in Pleše mining area
mines in Slovenia. Geologija, 63/2: 221–250. 
https://doi.org/10.5474/geologija.2020.018 
Gosar, M., Šajn, R., Miler, M., Markič, M. & Čar-
man, M. 2014: Izdelava popisa zaprtih objektov 
za ravnanje z odpadki iz rudarskih in drugih 
dejavnosti izkoriščanja mineralnih surovin: 
poročilo 3. faze projekta, elaborat, študija. 
Geološki zavod Slovenije, Ljubljana: 49 p.
Grunsky, E.C., Drew, L.J. & Sutphin. D.M. 
2009: Process recognition in multi-element 
soil and stream-sediment geochemical data. 
Appl Geochem, 24/8: 1602–1616. https://doi.
org/10.1016/j.apgeochem.2009.04.024
Hudson-Edwards, K.A., Macklin, M.G., Curtis, 
C.D. & Vaughan, D.J. 1996: Processes of
Formation and Distribution of Pb-, Zn-, Cd-,
and Cu-Bearing Minerals in the Tyne Basin,
Northeast England:  Implications for Metal-
Contaminated River Systems. Environ Sci
Technol, 30/1: 72–80. https://doi.org/10.1021/
es9500724
Levitan, D.M., Zipper, C.E., Donovan, P., 
Schreiber, M.E., Seal, R.R., Engle, M.A., 
Chermak, J.A., Bodnar, R.J., Johnson, D.K. & 
Aylor, J.A. 2015: Statistical analysis of soil ge-
ochemical data to identify pathfinders associ-
ated with mineral deposits: An example from 
the Coles Hill uranium deposit, Virginia, 
USA. J Geochem Explor, 154: 238–251. https://
doi.org/10.1016/j.gexplo.2014.12.012
Miler M., Bavec, Š. & Gosar, M. 2022: The en-
vironmental impact of historical Pb-Zn mi-
ning waste deposits in Slovenia. J Environ 
Manage, 308: 114580. https://doi.org/10.1016/j.
jenvman.2022.114580
Mlakar, I. 2003: On the problems of Ba, Pb, Zn 
Pleše ore deposit (Slovenia) = O problematiki 
Ba, Pb, Zn rudišča Pleše. Geologija, 46/2: 185–
224. https://doi.org/10.5474/geologija.2003.018
Mlakar, I. 1987: A contribution to the knowledge 
of the geological structure of Sava Folds and 
theirsouthern border – Summary = Prispevek 
k poznavanju geološke zgradbe Posavskih 
gub in njihovega južnega obrobja. Geologija, 
29/28: 157–182.
Placer, L. 1998: Structural meaning of the 
Sava folds = Strukturni pomen Posavskih 
gub. Geologija, 41/1: 191–221. https://doi.
org/10.5474/geologija.1998.012
Potra, A., Dodd, J.W. & Ruhl, L.S. 2017: 
Distribution of trace elements and Pb isotopes 
in stream sediments of the Tri-State mining 
district (Oklahoma, Kansas, and Missouri), 
USA, Appl Geochem, 82: 25–37. https://doi.
org/10.1016/j.apgeochem.2017.05.005
Premru, U. 1974: Triadni skladi v zgradbi osre-
dnjega dela Posavskih gub = Trias im ge-
ologischen Bau der mittleren Savefalten – 
Zusammenfassung. Geologija, 17: 261–297. 
Premru, U. 1980: Geological structure of Central 
Slovenia (Summary) = Geološka zgradba 
osrednje Slovenije. Geologija, 23/2: 227–278. 
Salminen, R., Batista, M.J., Bidovec, M., 
Demetriades, A., De Vivo, B., De Vos, W. et 
al. 2005: Geochemical atlas of Europe. Part 
1, Background information, methodology and 
maps. Geological Survey of Finland, Espoo: 
525 p.
Swan, A.R.H. & Sandilands, M. 1995: Introduction 
to Geological Data Analysis. Blackwell 
Science, Oxford: 446 p.
Teršič, T., Miler, M., Gaberšek, M. & Gosar, M. 
2018: Vsebnosti arzena in nekaterih drugih 
prvin v potočnih sedimentih in vodah pore-
čja Medije, osrednja Slovenija = Contents of 
arsenic and some other elements in stream 
sediments and waters of the Medija drainage 
basin, central Slovenia. Geologija, 61/1: 5–24. 
https://doi.org/10.5474/geologija.2018.001
TIBCO Software Inc. 2017: Statistica (data 
analysis software system), version 13. http://
statistica.io
Uradni list RS, 1996: Uredba o mejnih, opozoril-
nih in kritičnih imisijskih vrednostih nevar-
nih snovi v tleh. Ljubljana, Uradni list, 68: 
5773-5774. 
VROM, 2000: Dutch Target and Intervention 
Values (the New Dutch List), Circular on tar-
get values and intervention values for soil re-
mediation. The Ministry of Housing, Spatial 
Planning and the Environment (VROM). 
ht t p s : //e s d at . ne t / E nv i r on me nt a l%20
Standards/Dutch/annexS_I2000Dutch%20
Environmental%20Standards.pdf (Accessed 
16 August 2022).
Žebre, S.: Rudarska dejavnost v območju 
Posavskih gub. Rudarsko metalurški zbor-
nik, 1955,4: 239–255.
Žibret, G. & Čeplak, B. 2021: Distribution of Pb, 
Zn and Cd in stream and alluvial sediments 
in the area with past Zn smelting operations. 
Sci Rep., 11: 17629. https://doi.org/10.1038/
s41598-021-96989-y