Composition and origin of PM10 in Cape Verde: Characterization of long-range transport episodes

Accepted Manuscript Composition and origin of PM10 in Cape Verde: characterization of long-range transport episodes P. Salvador, S.M. Almeida, J. Cardoso, M. Almeida-Silva, T. Nunes, M. Cerqueira, C. Alves, M.A. Reis, P.C. Chaves, B. Artíñano, C. Pio PII:

S1352-2310(15)30639-7

DOI:

10.1016/j.atmosenv.2015.12.057

Reference:

AEA 14370

To appear in:

Atmospheric Environment

Received Date: 29 August 2015 Revised Date:

21 December 2015

Accepted Date: 23 December 2015

Please cite this article as: Salvador, P., Almeida, S.M., Cardoso, J., Almeida-Silva, M., Nunes, T., Cerqueira, M., Alves, C., Reis, M.A., Chaves, P.C., Artíñano, B., Pio, C., Composition and origin of PM10 in Cape Verde: characterization of long-range transport episodes, Atmospheric Environment (2016), doi: 10.1016/j.atmosenv.2015.12.057. This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

ACCEPTED MANUSCRIPT 1

Composition and origin of PM10 in Cape Verde: characterization of long-range

2

transport episodes

3

P. Salvadora, S. M. Almeidab,1, J. Cardosoc,d, M. Almeida-Silvab, T. Nunesc, M. Cerqueirac, C. Alvesc, M.

4

A. Reisb, P. C. Chavesb, B. Artíñanoa, C. Pioc

6

a

RI PT

5 Environmental Department of the Research Center for Energy, Environment and Technology (CIEMAT)

7

- Avenida Complutense 40, 28040 Madrid, Spain b

8

C2TN, Instituto Superior Técnico, Universidade Técnica de Lisboa, 2686-953 Sacavém, Portugal

d

10

CESAM, Aveiro University, 3810-193 Aveiro, Portugal

SC

c

9

Cape Verde University, Campus do Palmarejo, Praia, Cape Verde

M AN U

11 Abstract

13

A receptor modelling study was performed to identify source categories and their contributions to the

14

PM10 total mass at the Cape Verde archipelago. Trajectory statistical methods were also used to

15

characterize the main atmospheric circulation patterns causing the transport of air masses and to

16

geographically identify the main potential source areas of each PM10 source category. Our findings point

17

out that the variability of the PM10 levels at Cape Verde was prompted by the advections of African

18

mineral dust. The mineral dust load was mainly composed by clay-silicates mineral derived elements

19

(22% of the PM10 total mass on average) with lower amounts of carbonates (9%). A clear northward

20

gradient was observed in carbonates concentration that illustrates the differences in the composition

21

according to the source regions of mineral dust. Mineral dust was frequently linked to industrial emissions

22

from crude oil refineries, fertilizer industries as well as oil and coal power plants, located in the northern

23

and north-western coast of the African continent (29%). Sea salt was also registered in the PM10 mass

24

during most part of the sampling period, with a lower impact in the PM10 levels than the mineral dust one

25

(26%). Combustion aerosols (6%) reached the highest mean values in summer as a consequence of the

26

emissions from local-regional sources. Biomass burning aerosols produced from October to November in

27

sub-sahelian latitudes, had a clear influence in the content of elemental carbon (EC) recorded at Cape

28

Verde but a small impact in the PM10 total mass levels. A minor contribution to the PM10 mass has been

AC C

EP

TE D

12

1

Corresponding author Email: [email protected] IST/ITN, Instituto Superior Técnico, Universidade Técnica de Lisboa Estrada Nacional 10, ao km 139.7 2695-066 Bobadela LRS - Portugal Tel. +351-21-9946124 1

ACCEPTED MANUSCRIPT associated to secondary inorganic compounds-SIC. Namely, ammonium sulfate and nitrate (SIC 1 - 5%)

2

and calcium sulfate and nitrate (SIC 2 - 3%). The main origin of SIC 1 was attributed to emissions of SO2

3

and NOx from industrial sources located in the northern and north-western African coast and from

4

wildfires produced in the continent. SIC 2 had a clear regional origin in the summer period. However, in

5

the winter period there were probably contributions of soil emissions of evaporate minerals from regions

6

of eastern Algeria. The location of Cape Verde in the Atlantic Ocean at subtropical latitudes, and the

7

absence of relevant local sources of anthropogenic atmospheric pollutants, becomes this archipelago, a

8

perfect site to study the impact of external contributions on the background levels of PM10 registered over

9

the north-eastern tropical Atlantic.

RI PT

1

10

Keywords: Cape Verde; mineral dust; Sahara; source apportionment; trajectory statistical methods.

SC

11 12

1

Introduction

14

Mineral dust and sea salt are the largest sources of natural aerosols worldwide. Saharan desert is one of the

15

most important sources of mineral dust, contributing with more than 1900 million tons per year (Goudie,

16

2009) and being responsible for almost half of all the Aeolian material provided to the world’s oceans

17

(Miller et al., 2004). A persistent outflow of Saharan dust is transported to long distances over the

18

Mediterranean, Europe, North Atlantic Ocean and South America (Swap et al., 1992; Prospero, 1996;

19

Remoundaki et al., 2011; Salvador et al., 2014).

20

Mineral dust transport has implications on the local, regional and global climate and on environment

21

through the following processes: 1) direct effect on the shortwave and long-wave radiating flux through

22

scattering and absorption; 2) indirect influence on radiation budget through interfering in cloud formation;

23

3) semi-direct effect on relative humidity, vertical stability and precipitation and 4) affecting physical

24

parameters such as visibility (Goudie and Middleton, 2006; Klüser and Holzer-Popp, 2010; Knippertz and

25

Todd, 2012).

26

Saharan dust events exert significant effects upon air quality and, consequently, on human health and

27

well-being. Several epidemiological studies showed an association between atmospheric particulate matter

28

(PM) and an increase in morbidity and rate of mortality (Krewski et al., 2004; Samet and Krewski, 2007;

29

Almeida et al, 2014a; Cruz et al., 2015). According to the WHO (2013) new evidence suggests that short-

30

term exposures to coarse particles (including crustal material) are associated with adverse respiratory and

31

cardiovascular effects on health, including premature mortality. Desert dust episodes have been linked

32

with hospital admissions and mortality in a number of recent epidemiological studies. In Italy, the work

33

developed by Sajani et al. (2010) suggests an association between respiratory mortality in the elderly and

34

Saharan dust outbreaks. In Greece, Nastos et al. (2011) showed that during a Saharan dust episode the

AC C

EP

TE D

M AN U

13

2

ACCEPTED MANUSCRIPT hospital admissions due to respiratory diseases were 3-fold higher than the mean daily admissions. Pérez

2

et al. (2008) and Jiménez et al. (2010) found that increases in PM10 levels linked to inflows of Saharan

3

dust raised the risk of mortality in Barcelona and Madrid (Spain), respectively. More recently, Reyes et al.

4

(2014) found a significant increase in respiratory-cause hospital admissions in Madrid associated with

5

increases in PM10 and PM2.5-10 concentrations during African dust outbreaks.

6

It should also be noted that aside from mineral dust, anthropogenic pollutants (Rodríguez et al., 2011) and

7

microorganisms (Palmero et al., 2011) have been transported during Saharan dust events. Saharan storms

8

are thought to be responsible for spreading lethal meningitis spores throughout sub-saharan Africa, during

9

the dry season, where up to 250,000 people, particularly children, contract the disease each year and

RI PT

1

25,000 die (Pérez et al., 2014 and references therein).

11

Cape Verde (CV), an archipelago composed by 10 islands located offshore of western Africa coast, is

12

highly influenced by Saharan dust events (Chiapello et al., 1995; 1997). However, the Saharan dust impact

13

on the PM10 levels currently registered at CV has not been estimated before. Otherwise, at CV islands a

14

significant mixing of aerosols is expected due to the presence of sea salt and anthropogenic aerosol

15

emissions, aside from desert mineral dust. However, uncertainties exist in the quantification of the highly

16

variable distribution of this modified aerosol.

17

The main goal of this study was to investigate the influence of long-range transport episodes of PM on the

18

concentration levels and chemical composition of PM10 registered at CV, a remote background site located

19

in the north-eastern tropical Atlantic Ocean. This position in the Atlantic Ocean represents an important

20

area to study and characterize Saharan/Sahel mineral dust transported over west Africa and the adjacent

21

Tropical Atlantic Ocean.

22

In order to contribute for a better understanding and impact of the different sources, a 13-month

23

measurement campaign was performed at Santiago Island in the scope of the project “Atmospheric aerosol

24

in CV region: seasonal evaluation of composition, sources and transport” (CV-Dust) (Almeida-Silva et al,

25

2013, 2014; Fialho et al., 2014; Gonçalves et al., 2014). The main atmospheric circulation patterns causing

26

the transport of air masses at the synoptic scale were characterized by means of an objective classification

27

methodology of air mass back-trajectories arriving over CV. Then, the main PM10 sources contributing to

28

the levels registered at CV during the sampling period, were identified by the Positive Matrix

29

Factorization receptor model. Finally, the potential source areas of each PM10 source category were

30

identified by a specific trajectory statistical method, the Redistributed Concentration Field method, and

31

interpreted.

AC C

EP

TE D

M AN U

SC

10

32

3

ACCEPTED MANUSCRIPT 1

2

Materials and methods

2

2.1

3

A surface field station was implemented in the surroundings of Praia City at Santiago Island (14°55’ N

4

23°29’ W, 98 m asl), where aerosol sampling, with different samplers, was performed (Figure 1).

5

PM was collected simultaneously with two low-volume samplers (Tecora and Partisol) and one Hi-

6

Volume sampler (Tisch) all with standard PM10 inlet heads, between January 2011 and January 2012,

7

performing 140 parallel samples. The sampling time ranged between 6 to 96 h, decreasing during Saharan

8

dust episodes and increasing during periods with low PM concentrations in order to collect convenient

9

aerosol masses in the exposed filters. This procedure reduced the risk of filter clogging, decreased the

10

differences between masses collected in the different filters, benefited the chemical analysis, reduced the

11

measurement errors and equalized them among the different samples. Nuclepore polycarbonate filters

12

with 0.4 µm pore size, Teflon filters with 0.45 µm pore size and quartz filters were used with Tecora,

13

Partisol and Hi-volume samplers, respectively. Before and after sampling Nuclepore and Teflon filters

14

were stored in petri dishes and quartz filters were stored in aluminum foils. After sampling filters were

15

stored at a temperature of -20ºC until gravimetric and chemical analysis.

16

2.2

17

PM10 mass concentrations were quantified by gravimetric method with a micro balance (± 1 µg,

18

polycarbonate and teflon filters) and with an analytical balance (± 0.1 mg, quartz filter). The collected

19

filters were weighed in a controlled clean room (class 10,000) at 20 ± 1ºC and 50 ± 5% relative humidity,

20

and filters were kept for 24 hours in the same environment to equilibrate before weighing

21

(EN12341:1998). Filter mass before and after sampling was obtained as the mean of three measurements,

22

when observed variations were less than 1%.

23

Element analysis was carried out by Particle Induced X-Ray Emission (PIXE) and k0 - Instrumental

24

Neutron Activation Analysis (k0-INAA) at IST (Almeida et al, 2008). Each polycarbonate filter was

25

divided in four parts. For chemical identification one quarter was analyzed by PIXE - to measure the

26

elements Al, Br, Cr, Cu, Fe, K, Mn, Ni, Pb, Si, Ti, V and Zn - and another quarter by k0-INAA - to

27

determine As, Ce, Sb, Sc, and Sm.

28

PIXE analysis was carried out at a Van de Graaff accelerator, in vacuum and two X-ray spectra were taken

29

for each of the samples; one with a 1.2MeV proton beam and no absorber in front of the Si (Li) detector

30

for low energy X-ray elements and another with a 2.4MeV proton beam and a 250 µm Mylar® filter to

31

detect elements with atomic number higher than 20. The beam area at the target was 20 mm2.

32

For k0-INAA, the filter quarter was rolled up and put into a clean thin foil of aluminum and irradiated for 5

33

hours at a thermal neutron flux of 1.03×1013 cm-2.s-1 (f=103.4±1.3; α=-0.035±0.0001; Tn=330ºK) in the

AC C

EP

TE D

Gravimetric and Chemical Analysis

M AN U

SC

RI PT

Sampling

4

ACCEPTED MANUSCRIPT Portuguese Research Reactor (Dung et al., 2010). After irradiation, the sample was removed from the

2

aluminum foil and transferred to a polyethylene container. For each irradiated sample, two gamma spectra

3

were measured with a hyperpure germanium detector: one spectrum 2-3 days after the irradiation and the

4

other one after 4 weeks. Tests of reproducibility within the filters and between filters have been taken

5

previously, using parallel sampling with two similar sampling units and measuring the particle species by

6

k0-INAA. Results were reproducible to within 5–15%, providing strong support for the validity of the

7

analytical techniques (Almeida et al, 2003). The accuracy of analytical methods was evaluated with

8

certified reference materials, revealing results with an agreement of ±12% (Almeida et al., 2014b).

9

Teflon filters were used to quantify water inorganic soluble species by ion chromatography; while quartz

10

filter samples were used to quantify carbonaceous species. Carbonate fraction was determined by sample

11

acidification with phosphoric acid and the CO2 evolved measured with a NDIR gas analyzer. Elemental

12

and organic carbon were analyzed using a home-made thermal-optical system (Pio et al., 2011). Two 9

13

mm punches of each filter were used to determine the carbonaceous content. Controlled heating in anoxic

14

conditions (nitrogen atmosphere) until 600°C was performed to volatize the organic fraction, present in

15

the punch, followed by a second controlled heating phase, after cooling the oven to 350°C, in oxic

16

atmosphere (4% O2), up to 850°. The released carbon was oxidized to CO2 which was quantified

17

continuously by a non-dispersive infrared analyzer. Some organic compound is pyrolysed (PC) during the

18

first phase heating and was only released in oxidant conditions together with EC. The interference

19

between PC and EC was controlled by continuous evaluation of the blackening of filter using a laser beam

20

and a photodetector measuring the filter light transmittance. The released carbon until the moment when

21

the light transmittance recover it initial value was used to quantify OC and the remaining to the end of

22

heating gave the EC. This method was tested with the NIST 8785 filter standard and in an intercomparison

23

experiment with real aerosol samples (Schmid et al., 2001), delivering results between those obtained by

24

the NIOSH 5040, IMPROVE and EUSAAR2 protocols.

25

Blank filters were treated the same way as regular samples. All measured species were very

26

homogeneously distributed; therefore concentrations were corrected by subtracting the filter blank

27

contents.

28

2.3

29

2.3.1

30

Receptor modeling analysis with Positive Matrix Factorization (EPA-PMF 5.0) was employed to identify

31

the main sources of PM10 and estimate their contributions to the total mass (US-EPA, 2014).

32

Factor contributions and profiles are derived by the PMF model minimizing an objective function Q,

33

without detailed prior knowledge on source inventories (Paatero, 1999).

AC C

EP

TE D

M AN U

SC

RI PT

1

Statistical analysis

Positive Matrix Factorization (PMF)

5

ACCEPTED MANUSCRIPT Uncertainties used in the receptor modelling analysis were calculated based in Polissar et al. (2001). Only

2

elements with values of the signal-to-noise ratio (S/N) higher than 0.5, were selected for the present study

3

(US-EPA, 2014). This parameter indicates whether the variability in the measurements is real or within

4

the noise of the data. Some elements (Zn and Sc) that neither had a large influence on the chemical

5

profiles of the sources, nor a good correlation between observed and predicted values were excluded from

6

the final model solution. The model was repeteadly run and the number of sources was selected on the

7

basis of the variation of the objective function Q and the physical meaningfulness of the sources. Finally

8

the n=7 sollution was adopted.

9

Next, some tests were carried out to assess rotational ambiguity in the PMF solutions. The Base Model

10

Displacement Error Method was used to explore the rotational ambiguity in the PMF final solution. This

11

model assesses the largest range of source profile values without an appreciable increase in the Q value.

12

The adjustment in the factor profile values is always the maximum allowable, with the constraint that the

13

difference between the Q values associated with the original and the modified solutions (dQ) is not greater

14

than a specified value (dQmax). No factor swaps occurred for the smallest dQmax considered. It indicated

15

that there was not significant rotational ambiguity and that the solution was sufficiently robust to be used

16

(US-EPA, 2014).

17

Then, well known PM mass composition ratios were used to constrain the model run. Specifically, sea

18

salt (Na/Mg (m/m)=8.3, Na/k (m/m)= 27.6, Na/Ca (m/m)=25.85) and dust (Si/Na+=14.6, Si/Mg2+=198)

19

ratios were utilized. Besides, the mass ratio for ammonium sulfate SO42-/NH4+=2.67, was also considered.

20

No significant variations were observed for the source chemical profiles obtained for the constrained

21

model run solutions calculated for different dQ values (<5%) in comparison with the original solutions.

22

Rotational ambiguity were also assessed by examining scatter plots of one factor versus another using the

23

“G-Space Plot” option. Stable solutions were obtained, with zero contributions on both axes, in all the

24

cases.

25

2.3.2

26

In this study 5-day backward air trajectories arriving at the CV archipelago were used and analysed

27

together with the time series of PM10 concentrations and chemical composition recorded during the

28

measurement campaign. For each day of the sampling period four 3-D trajectories ending at 00:00, 06:00,

29

12:00 and 18:00 UTC over 16.85ºN, 24.87ºW were computed by the Norwegian Institute for Air Research

30

NILU using the FLEXTRA model and meteorological data provided from ECMWF. Trajectories arriving

31

at fixed heights of 500 and 1500 m asl were selected in this study to identify the main synoptic

32

meteorological situations that took place during the sampling period at CV.

33

In total, 1493 trajectories were computed for each arrival height, each with 40 endpoints (one time-step for

34

every 3-hours). A k-means cluster analysis (CA) was performed to group air mass back-trajectories into

EP

TE D

M AN U

SC

RI PT

1

AC C

Trajectory statistical methods (TSM)

6

ACCEPTED MANUSCRIPT similar groups, each one representing a characteristic meteorological scenario. The optimum number of

2

clusters to be retained was selected according to the percentage change in within-cluster variance as a

3

function of the number of clusters. Variance plots have been frequently used to identify features that

4

would indicate the correct number of clusters to be chosen for a given data set (Dorling et al., 1992;

5

Brankov et al., 1998).

6

Once the CA results were obtained, a characterisation of the meteorological scenarios causing the different

7

synoptic flows over CV was carried out. To this end, composite synoptic maps have been obtained by

8

averaging the sea level pressure and the geopotential height at the 850 hPa topography, using the data

9

corresponding to all days in which back-trajectories were assigned to a particular cluster in both data sets.

10

The meteorological variables used were obtained from the NCEP/NCAR Reanalysis datasets files (Kalnay

11

et al., 1996), provided by the NOAA/OAR/ESRL PSD, Boulder, Colorado, USA, from their Web site

12

(http://www.esrl.noaa.gov/psd/data/).

13

In this work the Redistributed Concentration Field (RCF) method (Stohl, 1996) was used to identify the

14

main potential source areas of the PM10 source categories identified with PMF. This method allows

15

simultaneous computational treatment of air mass back-trajectories and of time series of concentrations of

16

PM components registered at study sites. RCF obtained for well-known tracers of specific sources of PM

17

provides the location of the main polluting source areas contributing to these sites by medium- or long-

18

range transport. Back-trajectories arriving at 500 m and 1500 m asl were used with the aim to take into

19

account the influence of the air masses arriving at low and high altitude over CV with different wind

20

directions and speeds.

21

A 2º longitude x 2º latitude cells grid has been superimposed over the region defined by 2º-48ºN and

22

51ºW-21ºE. For each ij-th grid cell a weighted mean concentration was computed using the time series of

23

estimated contributions of the PM10 sources obtained with PMF and the data sets of back-trajectories time

24

steps. RCF results were reported on geographical maps as a result of the interpolation of the weighted

25

concentrations in the grid cells. Thus, areas with weighted concentrations in the higher and lower value

26

ranges indicated that, on average, air parcels residing over these cells resulted in high and low

27

concentrations, respectively, of the PM10 sources contributions at CV. It can thus be interpreted that the

28

prevailing synoptic winds, transported the PM10 source contributions from these potential source areas to

29

the receptor site. Hence, the interpretation of the RCF maps was achieved on the basis of the existence of

30

true emission sources over the potential source areas and the occurrence of specific atmospheric

31

circulation patterns that caused their transport. That was the reason why we performed a cluster analysis

32

for back-trajectories arriving at different heights and characterized the meteorological scenarios causing

33

the different synoptic flows over CV, associated to each cluster. It should be noted that regional

34

circulations, associated to the development of slow and moderate air flows cannot be represented by back-

35

trajectories occurring at relatively high altitudes (Stohl A., 1998).

AC C

EP

TE D

M AN U

SC

RI PT

1

7

ACCEPTED MANUSCRIPT It should be stressed that in spite of the fact that TSM holds great potential for identifying potential source

2

regions and preferred air mass transport pathways of PM components, they have been rarely used in

3

combination with other receptor modelling techniques, such as PMF, in PM source apportionment studies

4

(Belis et al., 2013). Poirot et al. (2001) recommended the application of multiple receptor techniques

5

(multivariate mathematical models as well as TSM) as a useful approach for improving the understanding

6

of source-receptor relationships for PM, for improving the confidence in the individual model results and

7

for developing a better understanding of the underlying aerosol data.

8

The RCF method has been recently used for identifying source areas of aerosol sources (Stohl, 1996; Lupu

9

and Maenhaut, 2002, Salvador et al., 2007; 2010; 2014, Kabashnikov et al., 2014), gaseous pollutants

10

(Scheifinger and Kaiser, 2007), acidic species in precipitation (Charron et al., 2000), dry deposition fluxes

11

(Han et al., 2004) and even isotopes in water vapour (Salamalikis et al., 2015). Polissar et al. (2001) used

12

PMF to identify the main sources of PM2.5 at Vermont (USA) and estimate their mass contributions.

13

Afterwards, they identified possible source areas by applying a TSM (potential source contribution

14

function-PSCF) to the daily source contributions identified by PMF. This method provides maps of

15

conditional probability values describing the spatial distribution of probable geographical source locations,

16

inferred by using trajectories arriving at the sampling site. Areas related to the high values of conditional

17

probability are interpreted as the potential source areas. Salvador et al. (2007) identified the main sources

18

of PM10 and PM2.5 at a background coastal site in Spain by means of a Varimax rotated factor analysis.

19

Then, they computed RCF for the elements that showed the higher factor loadings for each source, namely

20

the best source tracers. Scheifinger and Kaiser (2007) validated various TSM, including PSCF and RCF,

21

by different approaches, achieving the best performance with the RCF method. Besides, Wotawa and

22

Kröger (1999) successfully tested the ability of the RCF method to reproduce emission inventories of air

23

pollutants. For these reasons, we decided to compute RCF using the time series of PM10 source

24

contributions identified by PMF, as an innovative multicriteria approach to characterize the main sources

25

of PM10 at CV.

26

For any sample, the endpoints of all trajectories calculated during the sampling period, were combined

27

into a single composed trajectory. Those samples which were influenced by many different air mass

28

origins were not included in our analysis. With this aim a centroid was computed representing the

29

averaged flow of the group of back-trajectories for each sampling period. In order to assess quantitatively

30

the deviation of any single back-trajectory from the corresponding centroid, the Relative Horizontal

31

Transport Deviation (RHTD) (Stohl et al. 1998) between them was computed. When the RHTD for any

32

individual trajectory exceeded the 50% it was concluded that it has a substantial different origin than the

33

averaged flow represented by the centroid (Jorba et al. 2004). Therefore, composed trajectories, in which

34

more than 40% of individual trajectories were found to have substantially different origins, were discarded

35

from the trajectory analysis (Salvador et al., 2010). As a result, 10% and 14% of the aerosol samples were

36

excluded from the datasets used to create the RCF-500 and RCF-1500 maps, respectively.

AC C

EP

TE D

M AN U

SC

RI PT

1

8

ACCEPTED MANUSCRIPT From now on, the maps showing the RCF obtained with the back-trajectories arriving at 500 m and 1500

2

m asl will be referred to as RCF-500 and RCF-1500.

3

The location of the main focus with industrial activity in the central and northern African regions was

4

obtained from the Global Energy Observatory (http://globalenergyobservatory.org/) and online

5

information on chemical plants (http://www.afribiz.info/, http://www.wikipedia.com). See Figure 1 and

6

Supplementary Table 1 for details. This information was used to interpret the results obtained with the

7

CA, PMF and the RCF methods.

RI PT

1

8 9

3

Results and discussion

3.1

Characterization of the main atmospheric synoptic circulations over Cape Verde

11

According to the percentage change in within-cluster variance as a function of the number of clusters, the

12

optimum number of clusters retained was 7. The final cluster centers (average of the members of each

13

cluster) after the last iteration in the clustering procedure and the trajectories assigned to each cluster are

14

showed in Supplementary Figure 1. Both data sets represented similar air mass flows, but the cluster

15

centers obtained for the 500 data set, were in general shorter than for the 1500 one. As expected due to

16

their lower height, the 500 back-trajectories represents air mass movements with lower speeds and lenghts

17

than the air mass flows represented by the 1500 ones.

18

A marked seasonal pattern was observed in the occurrence of the different clusters (Figure 2). During the

19

summer, the most frequent air flows were those represented by clusters 1 and 2. These clusters were

20

constituted by slow regional and moderate north-westerly to north-easterly trajectories. They grouped 30-

21

33% of the total number of trajectories (Figure 2). The meteorological scenarios were characterized by the

22

presence of the sub-tropical “Azores high” pressure center at more western locations (35ºW) than usual

23

and a low baric gradient situation over CV (Figures 3a-b). In these conditions regional atmospheric

24

circulations over the CV archipelago prevails over fast advective air mass flows.

25

Fast North-African advection flows were the most frequent ones during the autumn and winter months

26

(Figure 2). Clusters 3 (7-12% of the total number of trajectories) and 4 (16-25%) gather fast north-easterly

27

and easterly trajectories of air mass, respectively, coming from the African mainland. The Azores high

28

was located over 25ºW, at higher latitudes in the case of cluster 3 than cluster 4, dominating the transport

29

of air mass over CV in this period (Figures 3c-d).

30

In the winter period, it was also detected the occurrence of fast north-westerly trajectories which were

31

grouped into cluster 5 (Figure 2). This synoptic meteorological situation was somewhat unusual (2-4% of

32

the total number of trajectories). It was produced owing to a strong longitudinal baric gradient generated

AC C

EP

TE D

M AN U

SC

10

9

ACCEPTED MANUSCRIPT over 30ºN, when the Icelandic low and the Azores high were shifted to the south of their normal positions

2

(Figure 3e).

3

The spring period was characterized by the advection of moderate western (cluster 6) and south-eastern

4

(cluster 7) flows (Figure 2). Cluster 6 grouped 8-12% whereas the trajectories that composed cluster 7

5

were the most frequent during the period of study (23-27% of the total). These synoptic situations were

6

characterized by a low baric gradient at tropical and sub-tropical latitudes (Figures 3f-g).

7 8 9

3.2

Variability of PM10 levels and chemical composition

RI PT

1

The average concentration levels of PM10 total mass and trace constituents are summarized in Table 1. The

11

average levels throughout this paper were weighted according to the sampling periods, with the aim to

12

reduce the influence of some highly intense events of transport of African dust. During these events the

13

highest PM10 concentration values of all the measurement campaign were recorded, during short sampling

14

periods.

15

Concentrations of PM10 experienced a large variability. The highest values were recorded in the winter

16

months (January-February and December) which are considered the “dusty season” at these latitudes

17

(Chiapello et al., 1997). In this period PM10 concentration values varied within the range 10.3-389.7 µg/m3

18

with a weighted mean concentration of 98.2 µg/m3. Increases in the PM10 mass were prompted by the

19

advections of African dust. This fact is supported by the simultaneous increases in PM10 and in Si, Ti, Al

20

and Fe concentrations registered at CV. Elements associated with clay-silicates (Sc, Ce, Cr, Al, Si, K, Mn,

21

Fe and Ti) exhibited a strong correlation (r2: 0.93 - 0.99) with PM10 (Supplementary Table S2). Moreover,

22

Si, Al and Fe presented high mean contributions to the total PM10 mass (11.0%, 6.4% and 3.1%

23

respectively, Table 1). During the winter period, CV is localized across the main path of African dust

24

transported by the north-eastern trade winds (Chiapello et al., 1995). The circulation patterns associated to

25

the clusters 3 and 4, describe very well the transport of African air masses towards CV, with high contents

26

of mineral dust. During the other seasons, the PM10 concentrations diminished abruptly, reaching weighted

27

mean values of 37.0, 41.0 and 64.6 µg/m3 in spring, summer and autumn, respectively. The lower

28

concentrations observed during summer at ground level at CV in comparison with those recorded in

29

winter, are due to the fact that much less mineral dust is carried by the trade winds in this period of the

30

year (Chiapello et al., 1995). As a consequence of the shift of the subtropical high towards the west, slow

31

and moderate air flows were frequently produced in summer driven by atmospheric circulations such as

32

those described in this work for the clusters 1 and 2. Hence, the mineral dust load in this period was

33

mainly constituted by contributions of local-regional dust. These contributions appeared all the year

AC C

EP

TE D

M AN U

SC

10

10

ACCEPTED MANUSCRIPT around in similar concentrations and mainly impacted in the concentration of coarse particles (Pio et al.,

2

2014).

3

Other components associated with mineral dust and with anthropogenic emissions, such as Ca2+, Mg2+,

4

CO32-, Sm, OC, Pb, Ni, Cu, As and Br, showed a moderate to high correlation (r2: 0.60 - 0.89) with PM10

5

(Supplementary Table S2). Finally, some components associated with secondary inorganic compounds –

6

SIC (SO42-, NO3-, NH4+), sea salt (Cl- and Na+) and anthropogenic emissions (V, Zn, Sb and EC) showed a

7

poor correlation, even negative (r2: (-0.32) - 0.42) with the PM10 levels (Supplementary Table S2). It

8

should be noted that apart from the crustal elements Si and Al, Cl- and Na+ which are currently associated

9

to marine spray, were the elements with the highest mean contribution to the total PM10 mass (9.0% and

RI PT

1

6.8%, respectively, Table 1).

11

Figure 4 shows the mean contribution of major elements to the PM10 mass, during events in which PM10

12

showed bulk mass concentration data lower than the 10th percentile (low PM10 events), within the 55th-80th

13

percentiles interval (moderate PM10 events) and higher than the 85th percentile (high PM10 events). All the

14

high PM10 events and 41% of the moderate ones were registered in winter. The low PM10 events were

15

mainly produced in spring (50%) and autumn (36%). It is evident that the high PM10 events were

16

characterized by an outstanding contribution of clay-silicates derived mineral compounds (Si, Al, Fe and

17

K). This type of mineral contribution was also significant during the occurrence of moderate PM10 events,

18

aside from the contribution of sea salt and of sulfate and nitrate compounds (Figure 4). Sea salt

19

contribution was the highest during low PM10 events, revealing the outstanding influence of the marine air

20

masses on the composition of the PM10 background levels at the CV archipelago.

21

The contribution of carbonaceous compounds and metals to the PM10 total mass, was very low during the

22

occurrence of any type of event. However, it was higher during low than medium and high PM10 events.

23

Supplementary Figure S2 demonstrates that during high PM10 events the concentration of most crustal

24

compounds (clay-silicates and calcite-dolomite derived mineral dust) and metals increased significantly

25

(>100%) with respect to their mean values. On the contrary, the mean values of these compounds during

26

moderate PM10 events were lower than during the whole sampling period. It represents the abrupt

27

increases of the PM10 levels during the African dust outbreaks, due to the high values of mineral dust load

28

transported and deposited over CV. Otherwise, the mean values of EC, Ca2+, sea salt (Cl- and Na+) and

29

SIC, increased discretely (from 10% to 40%) during moderate PM10 events with respect to the whole

30

sampling period. It indicates that other sources, apart from African dust, contributed to the concentration

31

levels of PM10 at CV but with less intensity.

AC C

EP

TE D

M AN U

SC

10

32

11

ACCEPTED MANUSCRIPT 3.3

Characterization of sources of PM10 at CV by PMF and RCF

2

PMF analysis allowed the identification of 7 main sources or origins for PM10 at CV. A good fitting

3

(R2=0.98) was obtained between predicted and experimental PM10 values.

4

The chemical profiles of the 7 sources are depicted in Supplementary Figure S3. The interpretation of the

5

sources was based on the presence of key tracers: Mineral 1 (Clay-silicates derived mineral dust with Al,

6

Ce, Fe, K, Mn, Si, Sm and Ti as main tracers), Mineral 2 (Carbonates: CO32- and Ca2+), Sea salt (Cl-, Na+,

7

K+ and Mg2+), Industrial+dust (anthropogenic industrial emissions mixed with mineral dust: Cr, Cu, Ni,

8

V, Pb, Sb, OC, Ti, Fe, K, Mn and Al), Combustion (EC and in a lesser extent OC, Sb and Cu), SIC 1

9

(NH4+, SO42- and NO3-) and SIC 2 (SO42-, NO3- and Ca2+). The weighted mean total and seasonal

RI PT

1

contribution of these sources to the PM10 mass were estimated and represented in Figure 5.

11

The Industrial+dust category was the most significant source contributing to the levels of PM10 at CV

12

(29% of the PM10 mass on average). The chemical profile obtained for this category showed that the mass

13

of Si, Al, Fe and OC represented 30%, 17%, 14% and 9% of the mean source contribution to the PM10

14

mass. The mass of some other industrial tracers such as Cr, Cu and Ni only represented 0.03-0.04% of the

15

mean source contribution. Hence, it can be concluded that the mineral dust load was higher than the

16

industrial contribution to the total PM10 mass. Sea salt and Mineral 1 contributions reached a weighted

17

average contribution of 26% and 22% of the PM10 mass, respectively. Contributions from Mineral 2 and

18

Combustion categories were less significant (9% and 6% of the PM10 mass, respectively) than from the

19

sources mentioned before. The lowest mean contribution was attributed to the SIC 1 and SIC 2 category

20

which only represented 5% and 3% of the PM10 mass, respectively.

21

On average, as already stated, the PM10 levels showed seasonal variations with higher values in winter

22

than in summer. In the winter season, the Industrial+dust and the Mineral 1 contributions reached by far

23

the highest values (38.0 and 41.4 µg/m3 on average) across all the seasons and categories. The highest

24

seasonal mean contribution of the Mineral 2 category was also obtained in winter (15.1 µg/m3 on

25

average). In this season, the long-range transport of air masses from northern and central Sahara was very

26

frequent, prompted by the atmospheric circulations described by the clusters 3 and 4. Otherwise, Sea salt

27

mean contribution reached similar mean values (12.7-15.4 µg/m3) in the winter, summer and the autumn

28

seasons. In fact, this was the highest contribution to the PM10 mass in the spring season (17.5 µg/m3 on

29

average) when the transport of pure maritime air masses (Cluster 6) took more frequently place.

30

In the summer season, the Combustion and the SIC 2 categories reached their highest contributions (6.3

31

and 2.2 µg/m3 on average). The SIC 2 mean contribution in the winter season (2.0 µg/m3) was also

32

relatively high. It could be attributed to the transport of evaporite minerals (anhydrite and gypsum) from

33

the Sahara desert in this period. SIC 1 contributions were higher along the spring, summer and autumn

34

seasons (3.4-4.4 µg/m3) than in the winter period (1.5 µg/m3).

AC C

EP

TE D

M AN U

SC

10

12

ACCEPTED MANUSCRIPT RCF-500 and RCF-1500 obtained for the Industrial+dust category, showed the main potential source

2

areas in the northern border line of Morocco and Algeria (Figure 6a-b). It is evident that this category

3

represents the mixing of dust with anthropogenic aerosols, originated from power plants and crude oil

4

refineries located in the northern and north-western border line of the mainland. The transport of polluted

5

air masses from this region to CV was attributed to north-eastern wind flows (Cluster 3). Alastuey et al.

6

(2005) and Rodriguez et al. (2011) demonstrated that African dust is very frequently mixed with North

7

African industrial pollutants and exported to the Canary Islands in the Saharan Air Layer.

8

The RCF-500 and RCF-1500 maps (Figure 6c-d) showed the main potential source areas of the Mineral1

9

category in central and southern Algeria and south-western Lybia. Mineral dust loading from these areas

10

are dominated by clay-silicates-derived elements, which are formed by Precambrian and Paleozoic massifs

11

(Moreno et al, 2006 and references therein). By contrast most of the northern and western Sahara lies upon

12

a carbonated lithology. The northward gradient over Algeria observed in the RCF-500 and RCF-1500

13

maps for the Mineral2 category (Figure 6e-f) demonstrated that dust collected in the northern Sahara

14

presents higher carbonates amounts than dust collected in the south Sahara and Sahelian regions, in good

15

agreement with previous studies focused on the soil composition of the source regions of African dust

16

(Chiapello et al., 1997). A higher CO32-/Si ratio was obtained for those samples recorded during prevailing

17

north-eastern (Cluster 3, mean ratio of 0.14) than eastern (Cluster 4; mean ratio of 0.08) advective

18

conditions.

19

The regions with the highest concentration values of Sea salt in the RCF-500 and RCF-1500 maps were

20

found as expected over the Atlantic Ocean (Figure 6g-h).

21

The RCF-500 map of the Combustion category (Figure 7a) showed as the main potential source area, the

22

regions over and surrounding the archipelago. Synoptic meteorological scenarios representing slow air

23

flows due to weak baric gradients occurred predominantly in the summer months. Such flows were

24

characterized by the back-trajectories arriving at the lower altitude (500 m asl) that composed the cluster

25

2. These scenarios favored the accumulation of the anthropogenic emissions produced from local and

26

regional sources. The origin of the Combustion emissions in the summer period was probably associated

27

to local-regional sources (traffic) as suggested by Gonçalves et al. (2014).

28

Otherwise, the RCF-1500 map identified a big area near the west coast of North Africa at 5-15º N and

29

extending from 15º W to 13º E, as the main potential source area of Combustion aerosols (Figure 7b).

30

Every year huge amounts of carbonaceous aerosols are produced from wildfires in Africa. From October

31

to April the most intense agricultural fires are registered in sub-Sahelian west Africa, (Ruellan et al., 1999;

32

Capes et al., 2008; Dall’Osto et al., 2010).

33

In this work we have accounted for the global distribution of TPM (total particulate matter) and BC (black

34

carbon) emissions from biomass burning using the Global Fire Emission Database version 3 (GFEDv3)

35

(Van der Werf et al., 2010; http://www.globalfiredata.org/). This database provides monthly emissions of

AC C

EP

TE D

M AN U

SC

RI PT

1

13

ACCEPTED MANUSCRIPT different atmospheric pollutants from wildfires on a global scale (0.5ºx0.5º). During the year 2011, high

2

emissions of TPM and BC from wildfires were detected across this area from January to April and from

3

October to December (Figure 8). The comparison between Figure 7b and Figure 8, suggests that biomass

4

burning carbonaceous aerosols originating from sub-Sahelian west African regions, were subject to

5

westward advection across the continent and the Atlantic Ocean. Due to the fact that the contribution of

6

the Combustion aerosols only reached 6% of the PM10 total mass (Figure 5), it should be expected a small

7

impact of the biomass emissions in the PM10 levels recorded at CV during the sampling period. The

8

circulation patterns obtained for clusters 4 and 7 (Figure 3), described the prevailing synoptic winds that

9

produced this long-range transport of carbonaceous aerosols to CV in the winter and the spring seasons.

10

The region centered over 10º N-2º E, (Figure 7b) was also identified by Capes et al. (2008) as an intense

11

source area of biomass burning emissions.

12

Otherwise, Gonçalves et al. (2014) determined the concentration values of monosaccharide anhydrides

13

(levoglucosan, mannosan and galactosan) in a group of 21 sets of the samples obtained at the CV

14

measurement campaign. Three to ten consecutive filters were pooled to perform the extractions. These

15

organic compounds are widely used as biomass burning makers. Their highest concentration values (>10

16

ng/m3), especially the levoglucosan ones (>8 ng/m3) were obtained in the sets of samples corresponding to

17

March, June and December 2011 (Gonçalves et al., 2014). It demonstrates that there was a small

18

contribution of biomass burning aerosols to the PM10 mass recorded at CV, probably with origin in the

19

continental wildfires during the spring and winter seasons and in local sources in the summer period.

20

The RCF-500 map for the SIC 1 category showed the region located over and nearby the industrialized

21

regions of Morocco and western Sahara coast (Figure 1) as the main potential source area (Figure 7c).

22

Emissions linked to oil refineries, phosphate-based fertilizer industry and power plants placed in north and

23

north-western African regions, were also identified by Rodriguez et al. (2011) as the most important

24

source of the NO3-, NH4+ and a fraction of SO42- observed in the Saharan Air Layer over the Canary

25

Islands. The occurrence of moderate (cluster 1) and fast (cluster 3) north-western air flows, in the summer

26

and the winter months (Figure 2), respectively produced the transport of these aerosols toward CV.

27

The presence of moderate amounts of Na+ and Mg2+ in the chemical profile of the SIC category

28

(Supplementary Figure S3) may indicate a prevailing reaction of the industrial emissions of SO2, H2SO4,

29

NOx and HNO3 with sea salt (ClNa and ClMg) and subsequent transport to CV (Harrison and Pio, 1983;

30

Pio and Lopes, 1998). The oceanic areas located at the west of the Morocco coast were also identified as a

31

potential source area of SIC 1, but less intense than the industrialized regions (Figure 7c). This could be

32

interpreted as the contribution of biogenic marine spray on the concentrations of sulfates.

33

The industrialized areas of the NW African coast were also identified as moderate source area of SIC 1 in

34

the RCF-1500 map (Figure 7d). However, the greatest potential contributions to SIC 1 concentrations were

35

located in sub-Sahelian west African regions, in good agreement with the result obtained for the

AC C

EP

TE D

M AN U

SC

RI PT

1

14

ACCEPTED MANUSCRIPT Combustion aerosols (Figure 7b) originated from biomass burning. In spite of the fact that the composition

2

of biomass burning aerosols is dominated by carbonaceous particles, inorganic species (SO42+, NO3+ and

3

NH4+) are also present in varying smaller quantities (Ruellan et al., 1999; Capes et al., 2008). Biomass

4

burning emissions from regions of west Africa south of the Sahara, are characterized by higher nitrate than

5

sulfate emissions (Savoie et al., 1989; Formenti et al., 2003). The NO3-/SO42-non-sea-salt ratios obtained

6

for samples recorded during prevailing eastern (Cluster 4, mean ratio of 1.18) and south-eastern (Cluster

7

7; mean ratio of 1.19) advective conditions, are consistent with that found in the trade wind aerosols

8

transported from Africa to Barbados in wintertime, attributed to biomass burning (ratio about 1.44, Savoie

9

et al., 1989).

RI PT

1

Finally, the RCF-500 map for the SIC 2 category suggested a prevailing local-regional origin in the

11

summer months (Figure 7e) also observed in the Combustion source case (Figure 7a). This source

12

category represents the contribution of particles of CaSO4 and Ca(NO3)2 to the PM10 mass registered at

13

CV. Their presence was likely due to the neutralization and attachment of gaseous emissions of SO2 and

14

NOx from local-regional sources with calcite (CaCO3) in soil particles (Pakkanen, 1996; Querol et al.,

15

1998).

16

Moreover, the RCF-1500 map indicates a probable transport of evaporite minerals (anhydrite and gypsum)

17

from the Bechar and Grand Erg Occidental regions in eastern and north-eastern Algeria, respectively

18

(Figure 7f). These regions are crossed by dry drainage systems and wadis that have been identified as

19

potential sources of evaporite minerals by a number of authors (Rodríguez et al., 2011; Ginoux et al.,

20

2012). This transport was probably produced in the winter months by intense north-eastern wind flows as

21

those described by the trajectories contained in the Cluster 3. It should also be noted that the Soralchir

22

Adrar oil refinery (Figure 1 and Supplementary Table 1) was located within the main potential source

23

showed in the RCF-1500 map (Figure 7c). The existence of particles of Ca(NO3)2 in this area could be due

24

to interaction of calcite with the NOx emissions from the oil refinery.

EP

TE D

M AN U

SC

10

AC C

25 26

4

Conclusions

27

In this study it has been demonstrated that depending on the origin of the air masses, the levels and

28

composition of the background aerosol registered at the Cape Verde location will be influenced by desert

29

mineral dust, mixed or free of the interference of industrial emissions, sea salt and biomass burning

30

emissions. The origin of the air masses varied seasonally and consequently PM10 concentration levels

31

experienced a large variability along the year. In the winter season intense advections of African dust

32

(clay-silicates and calcite-dolomite derived mineral dust) were frequently produced over the Cape Verde

33

islands, by the prevailing north-eastern trade winds. These events produced abrupt increases in the PM10

34

mass. It could be evidenced that when the air masses arriving at Cape Verde, resided previously over 15

ACCEPTED MANUSCRIPT industrial focus of atmospheric pollutants in the northern and northwestern coast of the African continent,

2

the mineral dust was very well mixed with industrial emissions originated from power plants, fertilizer

3

industries and crude oil refineries.

4

The advection of eastern and southeastern air masses from latitudes that varied from 0 to 15ºN, produced

5

the transport of biomass burning aerosols (mainly from sub-Sahelian West African regions during the

6

period October-April. This source, contributed to increase the concentration levels of EC, SO42-, NO3- and

7

NH4+ in PM10 at CV.

8

Sea salt contribution reached similar mean values along the seasons, reflecting the intense influence of the

9

marine air masses on the composition of the PM10 background levels over the Cape Verde archipelago.

RI PT

1

Long-range transport events of mineral dust and biomass burning emissions were less frequently produced

11

in the summer months, when slow and moderate air flows prevailed over the Cape Verde archipelago. In

12

this season the contributions of local-regional sources were the highest recorded along the sampling

13

period, giving rise to relatively high values of combustion aerosols (EC) and secondary inorganic

14

compounds (mainly CaSO4 and Ca(NO3)2). The specific atmospheric circulation patterns causing the long-

15

range transport events of atmospheric pollutants towards Cape Verde have been characterized. This

16

information can be used in forthcoming studies, aimed at distinguish properties of aerosols coming from

17

different origins.

M AN U

SC

10

18

Acknowledgements

20

This work was supported by the Portuguese Fundação para a Ciência e Tecnologia – FCT, under the

21

project CVDust: Atmospheric Aerosol in Cape Verde Region: Seasonal Evaluation of Composition,

22

Sources and Transport, PTDC/AAC-CLI/100331/2008. J. Cardoso and M. Almeida-Silva acknowledge

23

the FCT PhD grants SFRH/BD/6105/2009 and SFRH/BD/69700/2010, respectively and S.M. Almeida for

24

her contract IF/01078/2013. C2TN/IST authors gratefully acknowledge the FCT support through the

25

UID/Multi/04349/2013 project. ECMWF and NILU are acknowledged for providing the data sets and the

26

FLEXTRA trajectories computed from Cape Verde (http://www.nilu.no/projects/ccc/trajectories/). The

27

developers of the FLEXTRA model (Andreas Stohl, Gerhard Wotawa and Petra Seibert) are also

28

acknowledged. The authors would like to thank the NOAA/OAR/ESRL PSD, Boulder, Colorado, USA for

29

providing the meteorological dataset files (http://www.esrl.noaa.gov/psd/data/). The authors wish to thank

30

Dr. M. Pandolfi and Dr. F. Amato from the Institute of Environmental Assessment and Water Research

31

(IDAEA CSIC) for their assistance and the anonymous referee for his/her valuable comments.

AC C

EP

TE D

19

32 33

References 16

ACCEPTED MANUSCRIPT 1

Alastuey, A., Querol, X., Castillo, S., Escudero, M., Avila, A., Cuevas, E., Torres, C., Romero, P.-

2

M., Exposito, F., García, O., Pedro Diaz, J., Dingenen, R.V., Putaud, J.P., 2005. Characterisation of TSP

3

and PM2.5 at Izaña and Sta. Cruz de Tenerife (Canary Islands, Spain) during a Saharan Dust Episode (July

4

2002). Atmos. Environ. 39, 4715-4728.

5

of airborne particles. Nucl. Instrum. Methods Phys. Res. Sect. B 207, 434–446.

7 8

RI PT

6

Almeida, S.M., Reis, M.A., Freitas, M.C., Pio, C.A., 2003. Quality assurance in elemental analysis

Almeida, S.M., Freitas, M.C., Pio, C.A, 2008. Neutron Activation Analysis for Identification of African Mineral Dust Transport. J. Radioanal. Nucl. Chem. 276, 161-165.

Almeida, S.M., Silva, A.V., Sarmento, S., 2014a. Effects of exposure to particles and ozone on

10

hospital admissions for cardiorespiratory diseases in Setúbal, Portugal. J. Toxicol. Env. Health A 77, 837-

11

848.

SC

9

Almeida, S.M., Almeida-Silva, M., Galinha, C., Ramos, C.A., Lage, J., Canha, N., Silva, A.V.,

13

Bode, P., 2014b. Assessment of the Portuguese k0-INAA laboratory performance by evaluating internal

14

quality control data. J. Radioanal. Nucl. Chem. 300, 581-587.

M AN U

12

15

Almeida-Silva, M., Almeida, S.M., Freitas, M.C., Pio, C.A., Nunes, T., Cardoso, J., 2013. Impact

16

of Sahara dust transport on Cape Verde atmospheric element particles. J. Toxicol. Env. Health Part A 76

17

(4-5), 240-251.

Almeida-Silva, M., Almeida, S.M., Cardoso, J., Nunes, T., Reis, M.A., Chaves, P.C., Pio, C.A.,

19

2014. Characterization of the Aeolian aerosol from Cape Verde by k0-INAA and PIXE. J. Radioanal.

20

Nucl. Chem. 300, 629-635.

21

Brankov, E., Rao, S.T., Porter, P.S., 1998. A trajectory-clustering-correlation methodology for examining the long-range transport of air pollutants. Atmos. Environ. 32, 1525-1534.

EP

22

TE D

18

Belis, C.A., Karagulian, F., Larsen, B.R., Hopke, P.K., 2013. Critical review and metaanalysis of

24

ambient particulate matter source apportionment using receptor models in Europe. Atmos. Environ. 69,

25

94-108.

AC C

23

26

Capes, G., Johnson, B., McFiggans, G., Williams, P.I., Haywood, J., Coe, H., 2008. Aging of

27

biomass burning aerosols over West Africa: aircraft measurements of chemical composition,

28

microphysical

29

10.1029/2008JD009845.

properties,

and

emission

ratios.

J.

Geophys.

Res.

113,

D00C15,

doi:

30

Charron, A., Plaisance, H., Sauvage, S., Coddeville, P., Galloo, J.C., Guillermo, R., 2000. A study

31

of the source-receptor relationships influencing the acidity of precipitation collected at a rural site in

32

France. Atmos. Environ. 34, 3665-3674.

17

ACCEPTED MANUSCRIPT 1

Chiapello, I., Bergametti, G., Gomes, L., Chatenet, B., Dulac, F., Pimenta, J., Santos Suares, E.,

2

1995. An additional low layer transport of Sahelian and Saharan dust over the North-Eastern Tropical

3

Atlantic. Geophys. Res. Lett. 22, 3191-3194. Chiapello I., Bergametti G., Chatenet B., Bousquet P., Dulac F. and Santos Suares E., 1997.

5

Origins of African dust transported over the northeastern tropical Atlantic. J. Geophys. Res., 102, D12,

6

13701-13709.

RI PT

4

7

Cruz, A.M.J., Sarmento, S., Almeida, S.M., Silva, A.V., Alves, C., Freitas, M.C., Wolterbeek, H.,

8

2015. Association between atmospheric pollutants and hospital admissions in Lisbon. Environ. Sci. Pollut.

9

Res. 22 (7), 5500-5510.

Dall’Osto, M., Harrison, R.M., Highwood, E.J., O’Dowd, C., Ceburnis, D., Querol, X.,

11

Achterberg, E.P., 2010. Variation of the mixing state of Saharan dust particles with atmospheric transport.

12

Atmos. Environ. 44, 3135–3146.

SC

10

Dorling, S.R., Davies, T.D., Pierce, C.E., 1992. Cluster analysis: a technique for estimating the

14

synoptic meteorological controls on air and precipitation chemistry e method and applications. Atmos.

15

Environ. 26A, 2575-2581.

M AN U

13

16

Dung, H.M., Freitas, M.C., Santos, J.P., Marques, J.G., 2010. Re-characterization of irradiation

17

facilities for k0-NAA at RPI after conversion to LEU fuel and re-arrangement of core configuration, Nucl.

18

Instrum. Methods Phys. Res. Sect. A 622, 438-442.

Fialho, P., Cerqueira, M., Pio, C., Cardoso, J., Nunes, T., Custódio, D., Alves, C., Almeida, S.M.,

20

Almeida-Silva, M., Reis, M., Rocha F., 2014. The application of a multi-wavelength Aethalometer to

21

estimate iron dust and black carbon concentrations in the marine boundary layer of Cape Verde. Atmos.

22

Environ. 97, 136-143.

EP

TE D

19

Formenti, P., Elbert, W., Maenhaut, W., Haywood, J., and Andreae, M. O., 2003. Chemical

24

composition of mineral dust aerosol during the Saharan Dust Experiment (SHADE) airborne campaign in

25

the Cape Verde region, September 2000. J. Geophys. Res., 108(D18), 8576, doi:10.1029/2002JD002648.

AC C

23

26

Ginoux, P., Prospero, J. M., Gill, T.E., H. C. Hsu and M. Zhao, 2012. Global-scale attribution of

27

anthropogenic and natural dust sources and their emission rates based on MODIS Deep Blue aerosol

28

products. Rev. Geophys., 50, RG3005, doi:10.1029/2012RG000388.

29

Gonçalves, C., Alves, C., Nunes, T., Rocha, S., Cardoso, J., Cerqueira, M., Pio, C., Almeida,

30

S.M., Hillamo, R., Teinilä, K., 2014. Organic characterisation of PM10 in Cape Verde under Saharan dust

31

influxes. Atmos. Environ. 89, 425-432.

32 33

Goudie, A., Middleton, N., 2006. Desert dust in the global system. Germany. Springer. ISBN: 978-3-540-32354-9. 287 pp. 18

ACCEPTED MANUSCRIPT 1

Goudie, A., 2009. Dust storms: recent developments. J.Environ. Manage. 90, 89-94.

2

Han, Y., Holsen, T.M., Hopke, P.K., Cheong, J., Kim, H., Yi, S., 2004. Identification of source

3

locations for atmospheric dry deposition of heavy metals during yellowsand events in Seoul, Korea in

4

1998 using hybrid receptor models. Atmos. Environ. 38, 5353-5361.

5

aerosols of both marine and polluted continental origin. Atmos. Environ. 17, 1733–1738.

RI PT

6

Harrison, R.M., Pio, C.A., 1983. Size-differentiated composition of inorganic atmospheric

7

Jiménez, E., Linares, C., Martínez, D., Díaz, J., 2010. Role of Saharan dust in the relationship

8

between particulate matter and short-term daily mortality among the elderly in Madrid (Spain). Sci. Total

9

Environ. 408(23), 5729–5736.

11

Jorba, O., Pérez, C., Rocadenbosch, F., Baldasano, J.M., 2004. Cluster analysis of 4-day back

SC

10

trajectories arriving in the Barcelona area, Spain, from 1997 to 2002. J. Appl. Meteorol. 43, 887-901. Kabashnikov V., Milinevsky, G., Chaikovsky, A., Miatselskaya, N., Danylevsky, V., Aculinin, A.,

13

Kalinskaya, D., Korchemkina, E., Bovchaliuk, A., Pietruczuk, A., Sobolewsky, P., Bovchaliuk, V., 2014.

14

Localization of aerosol sources in East-European region by back-trajectory statistics. Int. J. Remote Sens.

15

35(19), 6993-7006.

M AN U

12

Kalnay, E., Kanamitsu, M., Kistler, R., Collins, W., Deaven, D., Gandin, L., Iredell, M., Saha, S.,

17

White, G., Woollen, J., Zhu, Y., Chelliah, M., Ebisuzaki, W., Higgins, W., Janowiak, J., Mo, KC.,

18

Ropelewski, C., Wang, J., Leetmaa, A., Reynolds, R., Jenne, R., Joseph, D., 1996. The NCEP/NCAR 40-

19

year reanalysis project. B. Am. Meteorol. Soc. 77, 437-470.

TE D

16

20 21

Klüser, L., Holzer-Popp, T., 2010. Relationship between mineral dust and cloud properties in the West African Sahel. Atmos. Chem. Phys. 10, 6901-6915.

23

Knippertz, P., Todd, M., 2012. Mineral dust aerosols over the Sahara: meteorological controls on

EP

22

emission and transport and implications for modeling. Rev. Geophys. 50, 1-28. Krewski, D., Burnett, R.T., Jerrett, M., Pope, A., Rainham, D.G., Calle, E.E., Thurston, G.D.,

25

Thun, M., 2004. Mortality and long-term exposure to ambient air pollution: Ongoing analyses based on

26

the American Cancer Society cohort. J. Toxicol. Env. Heal. A, 68, 1093-1109.

AC C

24

27

Lupu, A., Maenhaut, W., 2002. Application and comparison of two statistical trajectory

28

techniques for identification of source regions of atmospheric aerosol species. Atmos. Environ. 36, 5607-

29

5618.

30 31

Miller, R., Tegen, I., Perlwitz, J., 2004. Surface radiative forcing by soil dust aerosols and hydrologic cycle. J. Geophys. Res., 109, D04203, doi:10.1029/2003JD004085.

19

ACCEPTED MANUSCRIPT 1

Moreno, T., Querol, X., Castillo, S., Alastuey, A., Cuevas, E., Herrmann, L., Mohammed, M.,

2

Elvira, J., Gibbons, W., 2006. Geochemical variations in aeolian mineral particles from the Sahara–Sahel

3

Dust Corridor. Chemosphere 65, 261-270. Nastos, P.T., Kampanis, N.A., Giaouzaki, K.N., Agouridakis, P.I., atzarakis, A., 2011.

5

Environmental impacts on human health during a Saharan dust episode at Crete Island, Greece. Meteorol.

6

Z., 20, 517–529.

RI PT

4

7

Paatero, P., 1999. The multilinear engine – a table-driven, least squares program for solving

8

multilinear problems, including the n-way parallel factor analysis model. J. Comput. Graph. Stat. 8, 854-

9

888.

Palmero, D., Rodríguez, J. M., de Cara, M., Camacho, F., Iglesias, C., Tello, J. C., 2011. Fungal

11

microbiota from rain water and pathogenicity of Fusarium species isolated from atmospheric dust and

12

rainfall dust. J. Ind. Microbiol. Biot. 38, 13–20.

14

Pakkanen, T.A., 1996. Study of formation of coarse particle nitrate aerosol. Atmos. Environ. 30,

M AN U

13

SC

10

2475–2482.

15

Pérez, L., Tobías, A., Querol, X., Kunzli, N., Pey, J., Alastuey, A., Viana, M., Valero, N.,

16

González-Cabré, M., Sunyer, J., 2008. Coarse particles from Saharan Dust and Daily Mortality.

17

Epidemiology 19, 800–807, 2008.

Pérez, C., Stanton, M.C., Diggle, P.J., Trzaska, S., Miller, R.L., Perlwitz, J.P., Baldasano, J.M.,

19

Cuevas, E., Ceccato, P., Yaka, P., Thomson, M.C., 2014. Soil dust aerosols and wind as predictors of

20

seasonal

21

dx.doi.org/10.1289/ehp.1306640.

23

incidence

in

Niger.

Environ.

Health

Persp.

122,

679–686;

http://

Pio, C.A., Lopes, D.A., 1998. Chlorine loss from marine aerosol in a coastal atmosphere. J.

EP

22

meningitis

TE D

18

Geophys. Res. 103, 25263–25272.

Pio, C.A., Cerqueira, M., Harrison, R.M., Nunes, T., Mirante, F., Alves, C., Oliveira, C., de la

25

Campa, A.S., Artinano, B., Matos, M. (2011) OC/EC ratio observations in Europe: Re-thinking the

26

approach for apportionment between primary and secondary organic carbon. Atmos. Environ. 45 (34),

27

6121-6132.

AC C

24

28

Pio, C.A., Cardoso, J.G., Cerqueira, M.A., Calvo, A, Nunes, T.V., Alves, C.A., Custódio, D.,

29

Almeida, S.M., Almeida-Silva, M., 2014. Seasonal variability of aerossol concentration and size

30

distribution in Cape Verde using continuous aerossol optical spectrometer. Frontiers in Environmental

31

Science 2, 1-11

32 33

Polissar, A., Hopke, P., Poirot, R., 2001. Atmospheric aerosol over Vermont: Chemical composition and sources. Environ. Sci. Technol. 35, 4604–4621. 20

ACCEPTED MANUSCRIPT 1 2

Poirot, R., Wishinski P., Hopke, P., Polissar, A., 2001. Comparative application of multiple receptor methods to identify aerosol sources in Northern Vermont. Environ. Sci. Technol. 35, 4622–4636.

3

Prospero, J., 1996. Saharan dust transport over the north atlantic ocean and mediterranean: an

4

overview, in: Guerzoni, S., Chester, R. (Eds.), The Impact of Desert Dust Across the Mediterranean.

5

Environmental Science and Technology Library. Volume 11. ISBN: 978-90-481-4764-9, pp. 133-151. Querol, X., Alastuey, A., Lopez-Soler, A., Plana, F., Puicercus, J.A., Ruiz, C.R., Mantilla, E.,

7

Juan, R., 1998. Seasonal evolution of atmospheric suspended particles around a coalfired power station:

8

chemical characterisation. Atmos. Environ. 32 (4), 719–731.

10

Reyes, M., Diaz, J., Tobías, A., Montero, J. C., and Linares,C., 2014. Impact of Saharan dust particles on hospital admissions in Madrid (Spain). Int. J. Environ. Heal. R. 24, 63–72.

SC

9

RI PT

6

Remoundaki, E., Bourliva, A., Kokkalis, P., Mamouri, R., Papayannis, A., Grigoratos, T., et al.,

12

2011. PM10 composition during an intense Saharan dust transport event over Athens (Greece). Sci. Total

13

Environ. 409(20), 4361-4372.

M AN U

11

14

Rodríguez, S., Alastuey, A., Alonso-Pérez, S., Querol, X., Cuevas,E., Abreu-Afonso, J., Viana,

15

M., Pérez, N., Pandolfi, M., de la Rosa, J., 2011. Transport of desert dust mixed with North African

16

industrial pollutants in the subtropical Saharan Air Layer. Atmos. Chem. Phys. 11, 6663–6685. Ruellan, S., Cachier, H., Gaudichet, A., Masclet, P., Lacaux, J.P., 1999. Airborne aerosols over

18

central Africa during the experiment for regional sources and sinks of oxidants (EXPRESSO). J. Geophys.

19

Res. 104 (D23), 30673-30690.

TE D

17

Sajani, S. Z., Miglio, R., Bonasoni, P., Cristofanelli, P., Marinoni, A., Sartini, C., Goldoni, C. A.,

21

Girolamo, G., Lauriola, P., 2010. Saharan dust and daily mortality in Emilia-Romagna (Italy). Occup.

22

Environ. Med., 68, 446-451.

24

Salamalikis V., Argiriou A.A. , Dotsika E., 2015. Stable isotopic composition of atmospheric water vapor in Patras, Greece: A concentration weighted trajectory approach. Atmos. Res., 152, 93-104.

AC C

23

EP

20

25

Salvador, P., Artíñano, B., Querol, X., Alastuey, A., Costoya, M., 2007. Characterisation of local

26

and external contributions of atmospheric particulate matter at a background coastal site. Atmos. Environ.

27

41, 1-17.

28

Salvador, P., Artiñano, B., Pio, C., Afonso, J., Legrand, M., Puxbaum,H., Hammer, S., 2010.

29

Evaluation of aerosol sources at European high altitude background sites with trajectory statistical

30

methods. Atmos. Environ. 44, 2316–2329.

31

Salvador, P., Alonso-Pérez, S., Pey, J., Artíñano, B., de Bustos, J.J., Alastuey, A., Querol, X.,

32

2014. African dust outbreaks over the western Mediterranean Basin: 11-year characterization of

33

atmospheric circulation patterns and dust source areas. Atmos. Chem. Phys. 14, 6759-6775. 21

ACCEPTED MANUSCRIPT 2 3 4 5 6

Samet, J., Krewski, D., 2007. Health effects associated with exposure to ambient air pollution. J. Toxicol. Env. Health A 70(3-4), 227-242. Savoie, D.L., Prospero, J.M., Saltzman, E.S., 1989. Non-sea-salt sulfate and nitrate in trade winds aerosols at Barbados: Evidence for long-range transport. J. Geophys. Res. 94(D4), 5069-5080. Scheifinger, H., Kaiser, A., 2007. Validation of trajectory statistical methods. Atmos. Environ. 41, 8846-8856.

RI PT

1

Schmid, H., Laskus, L., Abraham, H.J., Baltensperger, U., Lavanchy, V., Bizjak, M., Burba, P.,

8

Cachier, H., Crow, D., Chow, J., Gnauk, T., Even, A., ten Brink, H.M., Giesen, K., Hitzenberger, R.,

9

Hueglin, C., Maenhaut, W., Pio, C. , Carvalho, A., Putaud, J.-P., Toom-Sauntry, D., Puxbaum, H., 2001.

10

Results of the “carbon conference” international aerosol carbon round robin test stage I. Atmos. Environ.

11

35, 2111–2121.

SC

7

Swap, R., Garstang, M., Greco, S., 1992. Saharan dust in Amazon Basin. Tellus, 44B, 133-149.

13

Stohl, A., 1996. Trajectory statistics-a new method to establish source-receptor relationships of air

14

pollutants and its application to the transport of particulate sulfate in Europe, Atmos. Environ. 30, 579-

15

587, 1996.

16 17

M AN U

12

Stohl A., 1998. Computation, accuracy and applications of trajectories - a review and bibliography. Atmos. Environ. 32, 947-966.

United States Environmental Protection Agency, 2014. EPA Positive Matrix Factorization (PMF)

19

5.0. Fundamentals and User Guide. Research and Development. EPA/600/R-14/108, April 2014,

20

www.epa.gov.

TE D

18

Van der Werf, G.R., Randerson, J.T., Giglio, L., Collatz, G.J., Mu, M., Kasibhatla, P.S.,

22

Morton,D.C., DeFries, R.S., Jin, Y., van Leeuwen, T.T., 2010. Global fire emissions and the contribution

23

of deforestation, savanna, forest, agricultural, and peat fires (1997–2009), Atmos. Chem. Phys., 10,

24

11707-11735, doi:10.5194/acp-10-11707-2010.

AC C

EP

21

25

Wotawa G., Kröger H., 1999. Testing the ability of trajectory statistics to reproduce emission

26

inventories of air pollutants in cases of negligible measurement and transport errors. Atmos. Environ., 33,

27

3037-3043.

28 29

WHO, Review of evidence on health aspects of air pollution – REVIHAAP Project: final technical report , World Health Organization, Regional Office for Europe, Copenhagen, 2013.

30

22

ACCEPTED MANUSCRIPT 1

Figure Captions: Figure 1. Location of industrial areas. See Supplementary Table S1 for details.

3

Figure 2.Percentage (a) and seasonal frequency of trajectories occurring in each cluster (b-h).

4

Figure 3. Composite 850 mb geopotential height (m) for the trajectories grouped in clusters 1-7.

5

Figure 4. Mean PM10 compounds levels during low, moderate and high PM10 events at CV.

6

Figure 5. Mean (a) total and seasonal (b) contribution of the sources obtained with PMF to the PM10 mass.

7

Figure 6. RCF for the PM10 sources “Industrial+Dust”, “Mineral 1”, “Mineral 2” and “Sea Salt” obtained

8

with PMF corresponding to back-trajectories arriving at 500 (left) and 1500 (right) m asl.

9

Figure 7. RCF for the PM10 sources “Combustion”, “SIC 1” and “SIC 2” obtained with PMF

SC

RI PT

2

corresponding to back-trajectories arriving at 500 (left) and 1500 (right) m asl.

11

Figure 8. Accumulated emissions of total particulate matter (TPM) a) and black-carbon (BC) b) from

12

wildfires (source: GFEDv3 database) in g/m2 burned, along the January-April and October-December

13

periods of 2011.

M AN U

10

14

Supplementary Figure S1. Cluster centers (average of the members of each cluster) resulting from the

16

analysis of the back-trajectories arriving at 500 (left) and 1500 (right) m asl.

17

Supplementary Figure S2. Variability in mean concentrations of major and trace components in PM10,

18

from the high and moderate PM10 events to the whole period.

19

Supplementary Figure S3. Source profiles resolved from PM10 samples analysed by PMF.

EP AC C

20

TE D

15

23

ACCEPTED MANUSCRIPT Table 1. Average levels and standard deviation (SD) of PM10 total mass and trace constituents (ng m-3). The mean contribution of each element to the PM10 total mass is also included. PM10 mass (%)

59602

92876

-

Al

3814

11508

6.40%

As

0.59

0.82

0.0010%

Br

8.3

5.9

0.014%

Ce

2.9

8.2

0.0049%

Cr

2.6

5.7

0.0044%

Cu

2.8

3.7

0.0047%

Fe

1835

4421

3.1%

K

772

1775

Mn

31

78

Ni

2.6

3.4

Sb

0.17

0.14

SC

PM10

1.3%

0.052%

0.0044%

M AN U

0.00030%

Zn

24

14

0.040%

EC

188

183

0.32%

Sc

0.44

1.2

0.00070%

Si

6595

18854

11%

Sm

0.35

0.89

0.00060%

197

441

0.33%

2.3

4.3

0.0039%

2.9

4.2

0.0049%

Ti V

TE D

Pb

OC

CO3

2-

Cl-

NO3-

2-

EP

SO4

Na

+

NH4

K

AC C

SD

RI PT

Average

+

+

965

1.6%

2610

1.4%

5344

2650

9.0%

1191

833

2.0%

1898

1384

3.2%

4047

2157

6.8%

213

301

0.36%

0.24

0.19

0.00040%

2+

386

216

0.65%

2+

818

1617

1.4%

Mg Ca

980

816

AC C

EP

TE D

M AN U

SC

RI PT

ACCEPTED MANUSCRIPT

AC C

EP

TE D

M AN U

SC

RI PT

ACCEPTED MANUSCRIPT

AC C

EP

TE D

M AN U

SC

RI PT

ACCEPTED MANUSCRIPT

AC C

EP

TE D

M AN U

SC

RI PT

ACCEPTED MANUSCRIPT

AC C

EP

TE D

M AN U

SC

RI PT

ACCEPTED MANUSCRIPT

AC C

EP

TE D

M AN U

SC

RI PT

ACCEPTED MANUSCRIPT

AC C

EP

TE D

M AN U

SC

RI PT

ACCEPTED MANUSCRIPT

AC C

EP

TE D

M AN U

SC

RI PT

ACCEPTED MANUSCRIPT

ACCEPTED MANUSCRIPT Highlights

PM10 sources at Cape Verde were characterized by multiple receptor techniques. PM10 levels variability was prompted by advections of African mineral dust. PM10 composition varied with the origin of the air masses.

RI PT

Mineral dust was frequently mixed with industrial emissions from northern Africa. Wildfires occurring at the African continent contributed to the levels of EC.

AC C

EP

TE D

M AN U

SC

Marine air masses strongly influenced the PM10 background levels.