Nuclear Instruments and Methods in Physics Research B 371 (2016) 376–380

Contents lists available at ScienceDirect

Nuclear Instruments and Methods in Physics Research B

journal homepage: www.elsevier.com/locate/nimb

Characterization and source apportionment of fine particulate

sources at Rijeka, Croatia from 2013 to 2015
Tatjana Ivošević a,⇑, Eduard Stelcer b, Ivica Orlić c, Iva Bogdanović Radović d, David Cohen b
a
Faculty of Engineering, University of Rijeka, Vukovarska 58, HR-51000 Rijeka, Croatia
b
Australian Nuclear Science and Technology Organisation, Locked Bag 2001, Kirrawee DC, NSW 2232, Australia
c
Department of Physics, University of Rijeka, Radmile Matejčić 2, HR-51000 Rijeka, Croatia
d
Laboratory for Ion Beam Interactions, Ruder Bošković Institute, Bijenička 54, HR-10000 Zagreb, Croatia

a r t i c l e i n f o a b s t r a c t

Article history: PM2.5 daily aerosol samples were collected in Rijeka, Croatia during period from 6th August 2013 to 29th
Received 27 June 2015 January 2015. In total, 259 samples were collected on Teflon filters and analyzed by PIXE and PIGE
Received in revised form 9 October 2015 techniques to give information on 21 elements from Na to Pb. Additionally, black carbon was determined
Accepted 12 October 2015
with the Laser Integrated Plate Method.
Available online 20 October 2015
Results were statistically evaluated using Positive Matrix Factorization (PMF). Eight major pollution
sources: auto, smoke, secondary sulfates, heavy oil combustion, sea spray, road dust, industry iron and
Keywords:
soil dust were identified together with their relative contributions in total PM2.5 pollution.
Aerosols
PM2.5
Ó 2015 Elsevier B.V. All rights reserved.
PIXE
PMF
Rijeka

1. Introduction
High concentrations of fine particles in the urban atmosphere
have adverse impact on air quality and consequently on human
health [1–3]. The identification and characterization of emission
sources and their contribution to the ambient concentration of
pollutants has been one of the major focus in urban air quality
research.
The city of Rijeka (latitude 45°210 N, longitude 14°260 E) is the
largest Croatian port, and the third city by size in the Republic of
Croatia with approximately 130,000 inhabitants. The industrial
complex, oil powered thermal plant 320 MW (TPP) and oil refinery
(OR), are located 9 km eastward from the city center. Other
possible pollution sources in this region are coal powered thermal
plant of 330 MW located at the Port Plomin (30 km southwest from
Rijeka) and industrial complex located in Trieste (60 km northwest
from Rijeka) with a very busy port, oil refinery and an 400 MW oil
power plant (Fig. 1).
Average daily wind speed was relatively low, averaging at the
1.6 m/s [4], which often caused accumulation of air pollution in
the bay of Rijeka.
During the 1.5 years sampling period (from August 2013 to
January 2015) 259 daily PM2.5 samples were collected. Samples
⇑ Corresponding author. Tel.: +385 91 932 10 05.
E-mail address: tatjana.ivosevic14@gmail.com (T. Ivošević).
were analyzed with Laser Integrated Plate Method (LIPM), Proton
Induced X-ray Emission (PIXE) and Proton Induced Gamma-ray
Emission (PIGE). The results obtained by LIPM, PIXE and PIGE
techniques were statistically evaluated using Positive Matrix
Factorization (PMF) with the intention to identify major sources
contributing to the fine particle emission.
2. Experimental
2.1. Sample collection
Sampling site was located in the Port of Rijeka, close to the main
bus stop and two busy roads with average traffic intensity of 25–30
thousand cars, trucks and buses per day.
PM2.5 samples were collected 4 times a week during the
18 months period (6th August 2013 to 29th January 2015). Due
to the technical problems with the sampler, samples were not col-
lected in the period from 5th May to 5th August 2014. The sampler
inlet was positioned 3 m above the sea level, 2 m above the ground
level and 30 m away from the main road. All samples were col-
lected during 24 h periods.
Cyclone sampler based on ANSTO ASP sampler [5] was used to
collect PM2.5 aerosols on a stretched Teflon filters (PALL
Corporation R2P1025, diameter of 25 mm, 3 lm pore size) with
the average flow rate of (22.5 ± 3) l/min. In total, 259 samples were
collected.

http://dx.doi.org/10.1016/j.nimb.2015.10.023
0168-583X/Ó 2015 Elsevier B.V. All rights reserved.
 T. Ivošević et al. / Nuclear Instruments and Methods in Physics Research B 371 (2016) 376–380 377

(Canberra, model SSL80155) placed at 145° for detection of more

energetic X-rays [6]. To optimize it for detection of higher energy
characteristic X-rays (>3 keV) a 360 lm thick Mylar film was put
in front of it to completely attenuate X-rays below 3.0 keV. Both
detectors were carefully calibrated using set of single element
Micromatter thin standards evaporated on thin Nucleopore
(polycarbonate) filters. One multielemental standard (Vienna Dust
Standard V98, Air particulate matter on filter media) was also mea-
sured. GUPIX software [7] was used for quantitative analysis of the
PIXE spectra and results were compared with the certified values.
At ANSTO, PIXE and PIGE techniques were used simultaneously
using 12 nA beam of 2.6 MeV protons with 8 mm diameter and col-
lection charge of 3 lC. Characteristic X-rays were measured with
SDD (165-VTX-EM) positioned at 145° for detection characteristic
X-rays. Measurements of absolute elemental concentrations in thin
samples were obtained through normalization to known thin
Fig. 1. Location of sampling site in Rijeka, thermal powered plants (TPP–oil and Micromatter reference material with typical areal density of
TPP–coal) and oil refinery (OR). approximately 50 lg/cm2 and quoted accuracy of ±5% [8]. The fol-
lowing six foils were used to cover the X-ray region from 1 keV to
20 keV: Al, Si, NaCl, CaF2 and SrF2. For calibration of PIGE detector
2.2. Analysis (Ge detector Canberra GC3020) Na and F reference foils were used.
The following elements were measured by PIXE: Al, Si, P, S, Cl, K,
The total PM2.5 mass (Table 1) was obtained by gravimetric Ca, Ti, V, Cr, Mn, Fe, Co, Ni, Cu, Zn, Br and Pb while PIGE was used
measurements with Mettler Toledo MX5 microbalance under the to measure only sodium.
laboratory conditions of 22 °C and relative humidity of 50%, with Typical uncertainties in measurements of concentrations by
uncertainty ±10 lg, readability 1 l, repeatability 0.0008 mg at PIXE (ANSTO and RBI) for major elements were from 6% to 8%
ANSTO. (S, K, Ca, Fe, Zn) and for minor elements were from 7% to 100%
All aerosol samples were analyzed by PIXE technique at two (Al, Si, P, Cl, Ti, V, Cr, Mn, Co, Ni, Cu, Br, Sr, Pb) [6,8]. Uncertainties
laboratories: Laboratory for Ion Beam Interactions, Ruder Bošković in concentrations of Na (PIGE and PIXE RBI) ranged from 6% to 18%.
Institute (RBI), Zagreb and Institute for Environmental Research To obtain black carbon (BC), Laser Integrating Plate Method
(ANSTO), Australia. Majority of samples (202) were analyzed at (LIPM) was used at the Laboratory for elemental microanalysis
RBI and only 57 samples at the ANSTO laboratory. (LEMA) in Rijeka, assuming a mass absorption coefficient of
At the RBI measurements were performed using 3 nA beam of 7.02 m2/g [9]. Uncertainties in concentrations of BC ranged from
1.6 MeV protons. Beam diameter at the target was 5 mm and total 7% to 10%.
collected charge was 3 lC. To measure wide range of elements,
two X-ray detectors were used: SDD (Ketek Vitus H20, 8 lm Be win- 3. Results and discussion
dow and 450 lm Si crystal thickness) placed at 150° for detection of
low energy characteristic X-rays (from Na to Fe) and a Si(Li) detector Results of PIXE and PIGE (ANSTO and RBI) are found to be in
very good agreement especially for elements with concentrations
above corresponding MDLs.
Table 1 In Fig. 2 presented is comparison between concentrations of Na
Average concentrations, standard deviations (SD), medians, maxima, minimum
detection limits (MDLs) in ng/m3 and the number N of samples above MDL (N) for
and Zn obtained by the two laboratories, ANSTO and RBI.
259 samples collected in Rijeka from period 6th August 2013 to 29th January 2015 Concentrations of Na obtained by PIXE at RBI were approximately
obtained by PIXE, PIGE and LIPM. 1% higher than corresponding concentrations obtained by PIGE at
El Avg SD Med Max MDL–RBI (ANSTO) N
ANSTO. Similarly, concentrations of Zn obtained at RBI were
approximately 1% lower than corresponding concentrations
Na 117 113 80.6 757 2.8 (62) 257
obtained at ANSTO. In general, concentrations obtained at these
Mg* 22 33 12 329 2.2 200
Al 44 77 22 650 2.2 (4.2) 255 two laboratories differed for approximately ±2%.
Si 110 196 54 1811 2.2 259 In Table 1, average concentrations of all measured elements,
P 2.8 6.9 1.2 93 3.6 (2.1) 103 their standard deviations, medians, maxima and minimum detec-
S 789 526 668 2975 4.4 (1.9) 259
tion limits (MDL) are given for all samples. In the last row we pre-
Cl 54 156 8 1445 2.5 228
K 194 278 116 2983 3.0 (1.2) 259 sented daily average, median and maximum of fine mass. It should
Ca 88 90 56 681 1.8 258 be noted that for the fine mass the EU Directive 2008/50/EC [10] is
Ti 3.4 4.7 1.9 34 2.5 253 defining only annual limit value of 25 lg/m3 (no daily limits are
V 3.4 4.7 2.5 48 0.4 (0.84) 231 given). As it can be seen from the Table 1, BC is dominant compo-
Cr 0.6 0.5 0.5 5.8 0.42 (0.78) 158
nent with the mean concentration of 3439 ng/m3, followed by S, K,
Mn 4.4 10 2.4 153 0.32 (0.62) 259
Fe 93 73 74 493 0.24 (0.56) 259 Na, Si, Fe, Ca and Cl which are indicating both anthropogenic and
Co 0.49 0.65 0.33 7.5 0.28 (0.59) 12 natural sources of PM2.5. To unfold key source fingerprints and
Ni 1.9 1.8 1.5 18 1.4 (1.8) 244 their contribution to total PM2.5 pollution PMF statistical method
Cu 4.4 11 2.6 150 0.23 (0.63) 259
was used as discussed below.
Zn 14 11 11 67 0.25 (0.48) 259
Br 2.6 1.9 2.2 19 0.2 (0.57) 218
Sr 1.5 5.7 0.39 75 0.96 45 3.1. PMF source profiles and contributions
Pb 6.8 25 4.0 404 2.3 (3.7) 217
BC 3439 1334 3357 8750 28 259 We applied Positive Matrix Factorization (PMF) to identify
PM2.5 20,600 8050 20,200 51,500
sources of fine particulates [11]. The original Paatero DOS version
*
202 samples. of the PMF codes (PMF2) was used [12]. The results of the PMF
378 T. Ivošević et al. / Nuclear Instruments and Methods in Physics Research B 371 (2016) 376–380

300

60
250
c_RBI / ngm-3

200
40

150

20

100 [Na]_RBI =1.01 [Na]_ANSTO +0.42 [Zn]_RBI = 0.993 [Zn]_ANSTO + 0.22

[Zn]_RBI =0.994[Zn]_ANSTO+0.22
2 2 2
R = 0.99 R = 0.98
R = 0.98
95% conf. int. 95% conf
95%. int.
conf. int.
50 0
50 100 150 200 250 300 0 20 40 60

c_ANSTO / ngm-3 c_ANSTO / ngm -3

Fig. 2. Comparison of Na and Zn concentrations obtained by the RBI and ANSTO laboratories.

1
0.1 Auto

0.01
0.001

1
0.1 Smoke

0.01
0.001
Relative contribution of elements of the source

1
Na Al Si P S Cl K Ca Ti V Cr Mn Fe Co Ni Cu Zn Br Sr Pb BC
0.1 Secondary sulfate
X Data
0.01
0.001

1
Na Al Si P S Cl K Ca Ti V Cr Mn Fe Co Ni Cu Zn Br Sr Pb BC
0.1 Heavy Oil comb
X Data
0.01
0.001

1
Na Al Si P S Cl K Ca Ti V Cr Mn Fe Co Ni Cu Zn Br Sr Pb BC
0.1 Sea spray
X Data
0.01
0.001

1
Na Al Si P S Cl K Ca Ti V Cr Mn Fe Co Ni Cu Zn Br Sr Pb BC
0.1 Road dust
X Data
0.01
0.001

1 Na Al Si P S Cl K Ca Ti V Cr Mn Fe Co Ni Cu Zn Br Sr Pb BC
0.1 Industry iron
X Data
0.01
0.001

1
Na Al Si P S Cl K Ca Ti V Cr Mn Fe Co Ni Cu Zn Br Sr Pb BC
0.1 Soil
X Data
0.01
0.001
0.0001
Na Al Si P S Cl K Ca Ti V Cr Mn Fe Co Ni Cu Zn Br Sr Pb BC

Elements

Fig. 3. Identification and characterization of eight sources of the fine particulate matter at Rijeka from 2013 to 2015.
 T. Ivošević et al. / Nuclear Instruments and Methods in Physics Research B 371 (2016) 376–380 379

Table 2 combustion, sea spray, road dust, industry iron and soil. Obtained
Average source contributions to the total fine mass given in percentage. fingerprints were attributed to the individual sources by taking
Factor/% Mass BC K S into account knowledge of local conditions and chemistry related
Auto 33 ± 3.5 46 1 4 to the composition of each source.
Smoke 21 ± 2.1 28 65 2 As mentioned earlier, the total number of analyzed samples was
Sec. sulfate 15 ± 2.5 1 5 81 259 out of which 18 samples were marked as ‘‘outliers”. After
Heavy oil com. 10 ± 2.7 12 3 8 removing outliers, the PMF matrix contained 241 samples and 20
Sea spray 9 ± 1.3 0 20 4
Road dust 6 ± 2.3 7 2 1
different elements (Na to Pb) plus BC. The best PMF results were
Industry iron 3.9 ± 2.1 6 1 0 obtained for 8 factors (Qtheor and Qexpt values were 2965 and
Soil 1.4 ± 1.0 0 3 0.3 3096.2 respectively, v2 was 1.04, and FPEAK value set to 0). The
PMF Fkey matrix input allows the user to influence the solution
by pulling down elements from identified source fingerprints,
analysis are shown on Fig. 3. Eight source fingerprints were identi- i.e. where a priori information is available that a particular element
fied. Each source was named based on the key representative should not be part of a source profile. For this study only BC was
elements in the source fingerprint and its relationships to the other pulled down to zero for the soil and sea spray. This was achieved
elements. These were: auto, smoke, secondary sulfate, heavy oil by specifying a value 6 in the Fkey matrix.

35
30
25 Auto
20
15
10
5
25
20 X DataSmoke
15
10
5

16
14
12 X Data Sec.sulfate
10
8
6
4
2
10
8 X DataHeavy Oil comb.
-3

6
c /µgm

4
2

25
20 X DataSea spray
15
10
5

14
12
10 X DataRoad dust
8
6
4
2
8
6 X DataIndustry iron
4
2

6
5
4 X DataSoil
3
2
1
0
Aug 2013 Jan 2014 May Aug Jan 2015

Date
Fig. 4. Time variations of PMF sources in Rijeka from 2013 to 2015.
380 T. Ivošević et al. / Nuclear Instruments and Methods in Physics Research B 371 (2016) 376–380
3.1.1. Factor 1 – auto
This factor is dominated by BC, Fe and S with traces of Cu, Zn,
Pb, K, Si, P, Cr, V and Ni. Iron, chromium and cooper are present
in this source with 70%, 68% and 60% of their total measured ele-
mental concentrations, respectively. The emission of Fe is typical
indicator of engine wear. Cu and Zn indicate breaks and tyre ware,
respectively. In this factor average V/Ni ratio is 0.9, which is typical
for diesel engine combustion [13] such as trucks, tractors and other
transport vehicles used for in-port transport. This factor represents
33% of the total fine particle mass (Table 2).
3.1.2. Factor 2 – smoke
Smoke from biomass burning is typically associated with high
concentration of K and BC. Potassium is present in this source with
65.5% of its total measured elemental concentration. This finger-
print also contains S, Fe, Zn and Pb. Only this factor shows strong
seasonal variations (Fig. 4) and contributes 21.1% of the total fine
particle mass (Table 2).
3.1.3. Factor 3 – secondary sulfate
This factor is dominated by S with traces of BC, K, Si, Na, Al and
Ca. Sulfur is present in this source with 80.9% of its total measured
elemental concentration. It represents 15.4% of the total PM2.5
mass.
3.1.4. Factor 4 – heavy oil combustion
Heavy oil combustion is mainly composed of BC, S and Na with
traces of Al, Si, K, V and Ni. Vanadium and nickel are present with
86.4% and 58.3% of their total measured elemental concentrations,
respectively and they are found to be highly correlated (r = 0.96).
Average V/Ni ratio in this source is found to be 2.7 which is indica-
tive for heavy oil combustion [14]. This ratio is typical of ship emis-
sions [15] and is represented by 9.9% of the total fine particle mass.
3.1.5. Factor 5 – sea spray
The main constituents in sea spray are Cl and Na with traces of
S, K, Ca and Br. Chlorine and sodium are present in this source with
99% and 85.6% of their total measured elemental concentrations,
respectively. The average Cl/Na mass ratio for sea water is 1.8
[16] while our finding ratio is 0.94. This suggests Cl depletion
(Cldep) or Na excess which is not associated with NaCl and sea
water (such as sodium nitrate) or both. This source represents
9.1% of the total fine particle mass.
3.1.6. Factor 6 – road dust
Road dust is dominated by BC and Ca with traces of S, Fe, K, Al,
Si and P. Calcium and phosphorus are present in this factor with
59.1% and 56.6% of their total measured elemental concentrations,
respectively. Calcium is from natural and anthropogenic origin,
56.7% of Ca belongs to calcium carbonate and 2.4% of Ca belongs
to calcium phosphate. Calcium phosphate is a fertilizer that is
occasionally offloaded in the Port of Rijeka. So, in this factor are
present elements from both, natural (Al, Si, Ca) and anthropogenic
sources (P, S, Cu, Zn, BC) and is named road dust [17]. This factor
represented 6% of the total PM2.5 mass (Table 2).
3.1.7. Factor 7 – industry iron
This factor is dominated by BC, Fe, Mn, K, Ca, Zn and Na.
Manganese is present in this source with 73.2% of its total
measured elemental concentration. Atmospheric Mn, Fe, Cu and
Zn are typically derived from steel industry [18]. Three km west
to our sampling site is one of the largest Croatian shipyards which
is most likely source of this factor. This source represents 3.9% of
the total fine particle mass.
3.1.8. Factor 8 – soil
Main constituents in soil factor are the typical crustal elements
like Al, Si, K, Ca, Ti, Fe. Silicon, aluminium and titanium are present
in this source with 62%, 54% and 40% of their total measured ele-
mental concentrations, respectively. Average Al/Si ratio is 0.38
which is common for alumina–silicates [19]. This source repre-
sented only 1.4% of total fine mass (Table 2). Measured elemental
ratios of soil dust agree with the composition of typical rocks for
this region which is not a typical quartz rock (Fig. 4).
4. Conclusion
Positive Matrix Factorization technique was used to determine
elemental source fingerprints and their relative contributions to
the total mass of fine particles collected in Rijeka, Croatia. The
results are based on 259 aerosol samples collected during the per-
iod from 6th August 2013 to 29th January 2015. PIGE, PIXE and
LIPM techniques were used to obtained concentrations, uncertain-
ties and MDLs for elements from Na to Pb, as well as BC. Major
sources of fine aerosol pollution in the region of Rijeka are found
to be: auto (33%), smoke (21.1%), secondary sulfates (15.4%), heavy
oil combustion (9.9%), sea spray (9.1%), road dust (6%), industry
iron (3.9%) and wind blown soil (1.4%). As expected, automobile,
biomass burning and industry are dominant sources of air
pollution in Rijeka.
Acknowledgements
The authors are grateful to Mr. V. Mezak from the Port of Rijeka
Authority, Mr. D. Kavre from the oil powered plant Rijeka,
Ms. M. Mioč and Mr. Z. Ciganj from the oil refinery Rijeka, Mr. D. Frka
and Ms. A. Tomljanović from the Road Authority Rijeka,
Mr. D. Mlinek from the National Meteorological and Hydrological
Service for their contribution to this work.
References
[1] D. Dockery, P. Stone, New Engl. J. Med. 356 (2007) 511.
[2] Churg, M. Brauer, M.C. Avila-Casado, T.I. Fortoul, J.L. Wright, Environ. Health
Perspect. 111 (2003) 714.
[3] C.A. Pope, R.T. Burnett, M.J. Thun, E.E. Calle, D. Krewski, K. Ito, G.D. Thurston, J.
Am. Med. Assoc. 287 (2002) 1132.
[4] Yearly report, <http://meteo.hr>.
[5] D.D. Cohen, E. Stelcer, D. Garton, J. Crawford, Atmos. Pollut. Res. 2 (2011) 182.
[6] M. Manousakas, E. Diapouli, H. Papaefthymiou, A. Migliori, A.G. Karydas, R.
Padilla-Alvarez, M. Bogovac, R.B. Kaiser, M. Jaksic, I. Bogdanovic-Radovic, K.
Eleftheriadis, Nucl. Instr. Meth. B 349 (2015) 114.
[7] J.L. Campbell, N.I. Boyd, N. Grassi, P. Bonnick, J.A. Maxwell, Nucl. Instr. Meth. B
268 (2010) 3356.
[8] D.D. Cohen, R. Siegele, I. Orlić, E. Stelcer, Nucl. Instr. Meth. B 189 (2002) 81.
[9] G. Taha, G.P. Box, D.D. Cohen, E. Stelcer, Aerosol Sci. Technol. 41 (2007) 266.
[10] eur-lex.europa.eu/LexUriServ/lexUriServ.do?uri=OJ:L:2008:152:0001:0044:
en:PDF.
[11] P.K. Hopke, Receptor Modelling in Environmental Chemistry, John Wiley &
Sons, New York, 1985.
[12] P. Paatero, U. Tapper, Environment 5 (1994) 111.
[13] D. Cesari, A. Genga, P. Ielpo, M. Siciliano, G. Mascolo, F.M. Grasso, D. Contini,
Sci. Total Environ. 497–498 (2014) 392.
[14] T. Moreno, X. Querol, A. Alastuey, Y.D. de la Rosa, A.M. Sanchez de la Campa, M.
C. Minguillon, M. Pandolfi, Y. Gonzales-Castanedo, E. Monfort, W. Gibbons, Sci.
Total Environ. 408 (20) (2010) 4569.
[15] M. Viana, P. Hammingh, A. Colette, X. Querol, B. Degraeuwe, I. de Vlieger, J. van
Aardenne, Atmos. Environ. 90 (2014) 96.
[16] D.R.E. Lide, Handbook for Chemistry and Physics, CRC Press, Boca Raton, FL,
1997.
[17] B.A. Begum, E. Kim, S.K. Biswas, P.K. Hopke, Atmos. Environ. 38 (2004) 3025.
[18] J.M. Pacyna, M.T. Scholtz, Y.-F. Li, Environ. Rev. 2 (1995) 145.
[19] D.D. Cohen, D. Garton, E. Stelcer, O. Hawas, T. Wang, S. Poon, J. Kim, B. Cheol
Choi, S.N. Oh, H.J. Shin, M.Y. Mi Young Ko, M. Uematsu, J. Geophys. Res. 109
(2004) 1.