INTRODUCTION Particulate matter, PM10 and PM2.5 is a serious problem for the

urban air pollution in many Bulgarian cities [1,2]. PM10 harmful

health effect is well known and depends on the concentration,

size and chemical composition.

In this study air particulate PM10 mass concentration and more

than 20 elements, measured by EDXRF technique, in the daily

quartz filter samples are obtained during winter and summer

experimental campaigns at NIMH observatory in Sofia. These

data and SO2 and NO2 concentrations from air quality urban

background stations are used to perform source apportionment

study. Multivariate techniques as Factor analyses (FA) and

Principle Components Analyses (PCA) by STATISTICA.6

software are applied to identify sources of PM10.

Application of PCA and FA models for source apportionment of PM10 in Sofia

Blagorodka Veleva1), Elena Hristova1), Maria Kolarova1), Emilia Nikolova2),

Raliza Valcheva2), Orlin Gueorguiev1)

1)National Institute of Meteorology and Hydrology, NIMH-BAS, Tsarigradsko sh. 66, Sofia, e-mail: elena.hristova@meteo.bg 2)Institute of Nuclear Research and Nuclear Engineering, INRNE-BAS, Tsarigradsko sh. 72, Sofia, e-mail: eminik@inrne.bas.bg

SUMMARY The ED-XRF technique was applied to determine more than 20 macro and micro elements in the PM10

filter samples in Sofia. The higher PM10 daily concentrations are related to the higher concentrations for

most of the elements, excluding Ag, Ca and Fe higher values in July.

The factors derived by PCA after varimax rotation are 3 to 6 depending of the selection of the set of

elements and should be considered as preliminary. More and enlarged data set is needed.

Further work is planned to determine water soluble ions in PM10 samples. EDXRF data from next

sampling campaigns will be utilized in further source apportionment study.

Map of Sofia and sampling site

15th EMS Annual Meeting & 12th European Conference on Applications of Meteorology (ECAM) , 07–11 September 2015, Sofia, Bulgaria

The whole data set (85 daily values) is divided in summer and winter subsets. The contribution of 5

factors for each group with different elemental components is derived. 85% of the PM10

concentration in summer is due to the impact of K, Ca, Ti, Fe source. In separate group with no

significant contribution are Zn and Cu.

When we use the entire data set, 51% of the contribution to PM10 is from sources related to the

traffic and road dust resuspension (NO2, Zn, Sr, Ti). The source of S and Mn (grouped in one factor)

contribute with 18% to the PM10 and the crustal elements source (K, Ca, Fe) with about 4.5%.

Median

25%-75%

Min-Max NO2 SO2 PM10

0E-01

2E+04

4E+04

6E+04

8E+04

1E+05

1E+05

1E+05

2E+05

2E+05

ng

.m-3

Factor analyses

The macro elements are found in the majority of the PM10 samples. The number of elements with

concentrations below Detection Limit (DL) is higher in summer. For example, Cd, I, Br, Zr and Ba are not

detected in the July 2012 samples when Pb, Rb and Ni were detected once. The As was detected only in

3 days in the summer of 2012 with maximum on 13 of July. Sn and Sb are below DL approximately in half

of the winter samples. Ni is detected in about 25% of total samples: measured once in July 2012 and in ~

50% of days in July 2013 . The macro elements K and Fe are in the range from less than 50 to 1000 ng.m-

3 in winter and from 100 to 1500 ng.m-3 in summer. In July, the daily Cu concentrations are several times

lower than in February, when the Cu concentrations are above 100 ng.m-3 for certain days. Similar are the

variations of Ti, Zn, Mn. Phosphorous shows high winter concentration, with average value of 534.1 - 590

ng.m-3 in 2012 and 2013 correspondingly, and of 31 – 500 ng.m-3 in July 2012 - 2013.

Win

ter

S

um

mer

The correlations between some elements and PM10 mass concentration are presented in scatter

plot figures for winter and summer experiments. Weak correlation between PM10 and Fe, K and Ti is

observed in winter periods. No correlation (r<0.3) is found between PM10and P, Ca, Mn, Sb and Ba.

Correlation coefficient > 0.6 is obtained between PM10 and NO2, Ti and Sr. The temporal variations

of these elements are influenced probably by different sources contribution and source profiles.

High seasonal variability of PM10/NO2 correlation coefficient is registered (0.85 in winter versus 0.18

in summer). Significant correlation (r>0.57) is obtained between elements pairs: K-Fe, K-Ca, K-Ba,

K-Ti, K-Cl, Ca-Fe, Ca-Ba, Sr-Ti, S-Mn.

PM10 mass concentration

Sulfur content in PM10 in winter is related to the same factor as Mn. It does not correlate significantly with most of the other elements in PM10 and NO2 and SO2 24-average

concentrations. High S concentrations and high S to PM10 ratio is registered in the PM10 samples of 6, 13, 21,22, 23 February 2012 and in 5 and 6 February 2013. HYSPLIT Trajectory

Model [3] is applied in cases with high sulfur concentration to trace back the origin of air masses. HYSPLIT dispersion model is run in case when back trajectories indicate the regional

transport from major industrial SO2 sources. In winter there is a local sulfur background because of the domestic heating.

Median

25%-75%

Min-Max P S Cl K Ca Fe

0

500

1000

1500

2000

2500

3000

ng

.m-3

Median

25%-75%

Min-Max Ti Mn Cu Zn Rb Sr Sn Sb Ba

0

20

40

60

80

100

120

140

160

180

200

ng

.m-3

NO2 vs. PM10

PM10 = -8766, + 1,2398 * NO2

Correlation: r = ,85525

0 20000 40000 60000 80000 1E5 1,2E5 1,4E5

NO2

-2E+04

0E-01

2E+04

4E+04

6E+04

8E+04

1E+05

1E+05

1E+05

2E+05

2E+05

PM

10

95% confidence

SO2 vs. PM10

PM10 = 23975, + 1,2328 * SO2

Correlation: r = ,47956

-10000 0 10000 20000 30000 40000 50000 60000

SO2

0,00E-01

2,00E+04

4,00E+04

6,00E+04

8,00E+04

1,00E+05

1,20E+05

1,40E+05

1,60E+05

1,80E+05

PM

10

95% confidence

Fe vs. PM10

PM10 = 18290, + 84,935 * Fe

Correlation: r = ,63027

-200 0 200 400 600 800 1000 1200

Fe

0E-01

2E+04

4E+04

6E+04

8E+04

1E+05

1E+05

1E+05

2E+05

2E+05

PM

10

95% confidence

Median

25%-75%

Min-Max PM10 NO2 SO2

0E-01

5E+03

1E+04

2E+04

2E+04

3E+04

3E+04

4E+04

4E+04

5E+04

ng

.m-3

Median

25%-75%

Min-Max P S Cl K Ca Fe

0

200

400

600

800

1000

1200

1400

1600

1800

ng

.m-3

NO2 vs. PM10

PM10 = 31969, - ,8256 * NO2

Correlation: r = -,1832

4E+03 5E+03 6E+03 7E+03 8E+03 9E+03 1E+04 1E+04 1E+04 1E+04

NO2

1E+04

2E+04

2E+04

3E+04

3E+04

4E+04

4E+04

5E+04

PM

10

95% confidence

SO2 vs. PM10

PM10 = 21890, + ,46350 * SO2

Correlation: r = ,21238

0E-01

2E+03

4E+03

6E+03

8E+03

1E+04

1E+04

1E+04

2E+04

2E+04

2E+04

SO2

1E+04

2E+04

2E+04

3E+04

3E+04

4E+04

4E+04

5E+04

PM

10

95% confidence

Fe vs. PM10

PM10 = 13035, + 21,424 * Fe

Correlation: r = ,84194

0 200 400 600 800 1000 1200 1400 1600

Fe

1E+04

2E+04

2E+04

3E+04

3E+04

4E+04

4E+04

5E+04

5E+04

PM

10

95% confidence

Su

mm

er

W

inte

r

Stagnant weather conditions and inversions are the major factor for the high PM10 values, registered

during the cold period. Higher PM10 daily concentrations correlate with higher concentrations of the S, K,

Ti, Sr. The PM10 mass concentration, the daily sum of precipitation and mixing height determined from 12

UTC aerological soundings is presented below.

Factor Loadings (Varimax normalized)_All Data set

(Marked loadings are >,700)

Factor 1 Factor 2 Factor 3 Factor 4 Factor 5

PM10 0,771 0,262 0,400 0,096 0,070

NO2 0,852 0,005 0,177 0,378 0,078

SO2 0,523 -0,096 0,563 0,171 -0,164

P 0,491 -0,224 -0,157 0,086 -0,630

S 0,142 0,156 0,844 -0,123 0,081

Cl 0,402 0,602 -0,231 0,504 -0,009

K 0,262 0,720 0,478 0,280 -0,006

Ca -0,055 0,915 -0,067 -0,011 -0,047

Ti 0,570 0,435 0,346 -0,083 -0,371

Mn -0,049 -0,036 0,792 0,236 0,234

Fe 0,103 0,909 0,048 -0,096 0,152

Cu 0,584 0,279 -0,197 0,017 -0,077

Zn 0,753 0,044 -0,009 0,129 0,384

Rb -0,120 -0,059 -0,258 -0,694 0,205

Sr 0,761 0,016 0,218 -0,419 -0,117

Sn 0,256 -0,058 0,153 0,095 0,822

Sb -0,016 0,091 -0,037 0,778 0,384

Ba 0,113 0,606 0,122 0,415 -0,019

Factor Loadings (Varimax normalized) _winter data set

(Marked loadings are >,700)

Factor 1 Factor 2 Factor 3 Factor 4 Factor 5 Factor 6

PM10 0,316 0,362 0,342 0,629 0,013 0,362

NO2 0,296 0,120 0,007 0,747 -0,079 0,433

SO2 -0,064 0,478 0,467 0,027 -0,180 0,505

P -0,042 0,113 -0,310 0,038 0,768 0,424

S 0,035 0,188 0,903 0,219 0,117 0,062

Cl 0,908 -0,007 -0,216 0,202 0,042 -0,031

K 0,781 0,112 0,528 0,131 0,117 0,086

Ca 0,883 -0,032 -0,033 0,046 0,201 0,018

Ti 0,324 0,252 0,097 0,029 0,065 0,866

Mn 0,028 -0,338 0,881 -0,082 -0,071 0,141

Fe 0,830 0,147 0,121 0,423 -0,012 0,049

Cu 0,354 0,686 -0,189 0,271 0,017 0,058

Zn 0,086 0,011 -0,043 0,856 0,073 0,071

Rb -0,161 -0,075 -0,391 0,042 -0,813 0,126

Sr -0,102 -0,005 0,136 0,464 0,255 0,839

Sn -0,067 -0,134 0,295 0,635 -0,477 -0,258

Sb 0,305 -0,758 -0,056 0,125 -0,298 -0,256

Ba 0,741 -0,067 0,096 -0,143 -0,167 0,175

Backward trajectories for

13.02.2012 sources from Serbia

could contribute to the elevated S

concentration. Dispersion model is

run for TPP “Nikola Tesla A” and

Mining and Smelting Basin Bor

High S concentration in PM10 during

the period 21-23 February 2012

most probably is due to Bulgarian

sources (TPP Maritza Iztok, Arubis

and TPP Republika, Pernik)

Factor Loadings (Varimax normalized) summer

data set, (Marked loadings are >,700)

Factor 1 Factor 2 Factor 3 Factor 4

PM10 0,958 0,064 0,000 -0,173

NO2 -0,230 -0,447 0,595 -0,126

SO2 0,254 0,191 0,650 0,332

S 0,640 -0,243 -0,160 -0,381

K 0,925 0,198 0,105 0,025

Ca 0,856 0,047 0,329 0,097

Ti 0,897 0,200 0,032 0,078

Mn 0,026 -0,257 -0,111 0,823

Fe 0,905 0,102 0,203 0,113

Cu -0,054 -0,822 0,133 0,026

Zn -0,144 -0,827 -0,146 0,250

Rb -0,292 0,038 -0,851 0,323

Element concentrations

TPP “Maritza Iztok” Aurubis - Copper and Copper Recycling TPP “Republika”, Pernik TPP “Nikola Tesla A” Mining and Smelting Basin Bor

Median

25%-75%

Min-Max Ti Mn Cu Zn Rb Sr Sn Sb Ba

0

20

40

60

80

100

120

ng

.m-3

0

100

200

300

400

500

600

700

0

20

40

60

80

100

120

140

160

180

06

.2.1

2

07

.2.1

2

08

.2.1

2

09

.2.1

2

10

.2.1

2

11

.2.1

2

12

.2.1

2

13

.2.1

2

14

.2.1

2

15

.2.1

2

16

.2.1

2

17

.2.1

2

18

.2.1

2

19

.2.1

2

20

.2.1

2

21

.2.1

2

22

.2.1

2

23

.2.1

2

24

.2.1

2

25

.2.1

2

H(m)PM10

mg

.m-3

, m

m

PM10 Precip. Hmix

0

100

200

300

400

500

600

700

0

10

20

30

40

50

60

70

80

90

21

.1.1

3

22

.1.1

3

23

.1.1

3

24

.1.1

3

25

.1.1

3

26

.1.1

3

27

.1.1

3

28

.1.1

3

29

.1.1

3

30

.1.1

3

31

.1.1

3

01

.2.1

3

02

.2.1

3

03

.2.1

3

04

.2.1

3

05

.2.1

3

06

.2.1

3

07

.2.1

3

08

.2.1

3

09

.2.1

3

10

.2.1

3

11

.2.1

3

H(m)PM10

mg

.m-3

, m

m

PM10 Precip. Hmix

0

200

400

600

800

1000

1200

1400

1600

1800

2000

0

5

10

15

20

25

30

35

40

45

02

.7.1

2

03

.7.1

2

04

.7.1

2

05

.7.1

2

06

.7.1

2

07

.7.1

2

08

.7.1

2

09

.7.1

2

10

.7.1

2

11

.7.1

2

12

.7.1

2

13

.7.1

2

14

.7.1

2

15

.7.1

2

16

.7.1

2

17

.7.1

2

18

.7.1

2

19

.7.1

2

20

.7.1

2

21

.7.1

2

22

.7.1

2

23

.7.1

2

24

.7.1

2

25

.7.1

2

26

.7.1

2

H(m)PM10

mg

.m-3

, m

m

PM10 Precip. Hmix

0

200

400

600

800

1000

1200

1400

1600

1800

2000

0

5

10

15

20

25

30

35

40

45

01

.7.1

3

02

.7.1

3

03

.7.1

3

04

.7.1

3

05

.7.1

3

06

.7.1

3

07

.7.1

3

08

.7.1

3

09

.7.1

3

10

.7.1

3

11

.7.1

3

12

.7.1

3

13

.7.1

3

14

.7.1

3

15

.7.1

3

16

.7.1

3

17

.7.1

3

18

.7.1

3

19

.7.1

3

20

.7.1

3

21

.7.1

3

22

.7.1

3

23

.7.1

3

H(m)PM10

mg

.m-3

, m

m

PM10 Precip. Hmix

SAMPLING AND METHODS •PM10 sampling was carried out at the site of Central

Meteorological Station (CMS) of NIMH

•Sampling periods – 6.02–25.02.2012, 2.07-

26.07.2012, 21.01-11.02.2013, 1-23.07.2013.

•Certified PM10 sampler TECORA Echo PM is used.

•Sampling duration is 24h (at 8:00 LST).

•47mm quartz fiber filters (Whatman ). Filters were

conditioned for 48h before and after sampling in a

temperature and humidity controlled room (T=

20±2°C, RH=50±5%).

•Energy Dispersive X-Ray Fluorescence (EDXRF)

technic in INRNE is used .

ACKNOWLEDGEMENTS

This presentation was supported by the Project

BG051PO001-3.3.06-0063 funded by European

Operational Program “Human Resources

Development”.

References:

[1] Nat. Rep. of MoEW, (2013) National Report on the State and Protection of The Environment, MoEW, Sofia,

http://eea.government.bg/bg/soer/2013 ;

[2] Hristova, E., Veleva, B. (2013) ‘Variation of air particulate concentration in Sofia, 2005-2012’, Bulgarian J. of Met. and

Hydr., Vol. 18, No1-2, pp 47-56.;

[3] Draxler, R.R. and Rolph, G.D. (2013) ‘HYSPLIT (HYbrid Single-Particle Lagrangian Integrated Trajectory) Model

access via NOAA ARL READY Website (http://www.arl.noaa.gov/HYSPLIT.php). NOAA Air Resources Laboratory,

College Park, MD.

Statistical analysis