1 Introduction

Suspended particulate matter (PM) is one of the pollutants which require monitoring as a component of the strategy to manage the risk motivated by its health, ecological and civil impacts (Molnár and Mészáros 2001; Reis et al. 2002; Bates et al. 2005; Künzli et al. 2005; EPA 2006). Major sources of aerosols are heterogeneous in size and composition (Pacyna 1986), making it possible to attribute each fraction of the mass collected on a filter to a suspected source using knowledge of it’s chemical composition. Furthermore, if there are several contributors of specific chemical species, the relative load of each chemical species within the source emission may still allow the scientist to identify the individual sources in a single aerosol sample (Li et al. 2003). Receptor modelling techniques are useful tools for identifying individual sources in a sample of atmospheric pollution.

The analysis of the sources contributing to PM levels in Cartagena is especially interesting given the complex mixture of anthropogenic and natural emissions in the area. Figure 1 shows the locations of the major sources in the Cartagena area including (1) harbour activities, (2) a petrochemical factory, (3) a plastic production factory, (4) a power station (sometimes assisted by oil combustion), (5) a Zn metallurgical industry, (6) a military shipyard, (7) dust outbreaks from the northern African deserts, and (8) resuspension of local and/or regional particulate matter from a variety of origins. The major sources of resuspended PM include: (1) soils with variable Pb, Zn, Cd or Cu content in the vicinity of the city (Martínez García et al. 2001) under the influence of the persistent aridity, this is due to a climate with mild temperature and scarce precipitation during the year and extremely low precipitation in summer, when temperature and insolation are maximal (Capel Molina 1986); (2) agriculture in the surrounding area named “El Campo de Cartagena”; and (3) resuspension of city dust (mainly from construction and demolition activities).

Fig. 1
figure 1

Plot of the main locations in the survey. 1 Petrochemical factory, 2 plastic production factory, 3 power station, 4 zinc metallurgical industry, 5 military shipyard, 6 city dock, 7 Escombreras dock; A PM2.5 Santa Ana sampling station, B PM10 Bastarreche sampling station

Full size image

In 1979 Cartagena was declared “area of atmospheric pollution”, this term supposed a series of legal and subsequent technical measures in order to reduce the concentration of sulphur dioxide and suspended particulate matter; these pollutants were overcoming their thresholds in a way enough to motivate the mentioned response from the Spanish Government (Decreto 833/1975, 6 February. BOE-Official Spanish Gazette- 96, 4/22/1975 and Real Decreto 1197/1979, 20 February. BOE-Official Spanish Gazette-123, 5/23/1979). Although the situation had improved by the early 80s, the air quality was still undesirable (Moreno-Grau 2006). Harbour activities including the unloading of soybean gave rise to several epidemic pollution episodes, with the result of one fatality during an episode in 1987 (Consejería de Sanidad de la Región de Murcia 1991). As a consequence of these episodes, the regional authorities implemented a strategic plan which reduces or stops factory emissions during meteorological conditions associated with elevated pollutant levels in the Cartagena atmosphere (Cirera 2005).

The implementation of air quality measures should pay attention to source contribution analysis, as it points to the contributing sources which should be addressed (Querol et al. 2001a). Source contributing studies can establish a robust knowledge base needed for the design of the policies or strategies that minimize the impact of anthropogenic emissions and identify the episodes with a natural origin. As part of the effort to implement the assessment and control of the atmospheric pollution in Cartagena, this paper characterizes the complexity of the aerosol and applies statistical models for identifying the multiple sources.

2 Materials and Methods

2.1 Experimental

Two aerosol monitoring locations belonging to the Cartagena Local Council were chosen for this survey: Bastarreche and Santa Ana (Fig. 1). PM10 sampling (DIGITEL DHA-80, Switzerland) was performed at the hotspot traffic monitoring station at “Bastarreche Square”, an entrance from a highway to the city, near the coach and train stations (20 m a.s.l). This sampling lasted from February 25, 2004 to March 15, 2005 with a total of 98 samples were collected on quartz micro-fibre filters (Schleicher & Schuell QF20, Germany). PM2.5 sampling was carried out at the suburban monitoring station at “Santa Ana”, a small village around 3 km from Cartagena. Sampling at this site lasted from January 16, 2004 to March 2, 2005 with a total of 98 samples collected with the same filters and sampler.

Once sampled, the filters were sent to laboratory to determine PM ambient air concentrations, according to the procedure described in Querol et al. (2004a,b): The methodology included the successful performance of an IEA round robin exercise, the use of reference material in keeping with relative analytical errors lower than 10% for most elements. The analytical methods include (a) PM levels were determined by gravimetry; (b) Major and trace chemical species determined by acid digestion and further analysis by ICP-MS, and ICP-AES; (c) Soluble anions were determined by water leaching and analysis by ion chromatography, and ammonium by FIA colorimetry; and (d) Organic matter and elemental carbon (OM+EC) were determined by applying a factor of 1.2 to the total C content determined in a CHNS elemental analyser (Turpin et al. 2000).

2.2 Statistical Treatment and Source Contribution Method

One approach to receptor modelling is based on factor analysis (FA). The FA applied in this study is principal component analysis (PCA; Thurston and Spengler 1985). The results obtained with this analysis were then interpreted using complementary information from meteorological data and back-trajectory analysis (Draxler and Hess 2004), satellite imagery (McClain et al. 1998), as well as support from regional dust modelling (Kallos et al. 1997; Nickovic et al. 2001; NRL 2003). The purpose of FA is to reduce the number of dimensions in data sets containing a large number of variables in order to facilitate the interpretation of the observed aerosol (Hopke 1992). To aid in the interpretation of the main common factors, the PCA method was applied with a varimax rotation to the set of data of the chemical species. The result was the production of a number of independent factors that maximize and minimize the associations between individual variables and factors. Each factor can be described as a linear combination of the studied variables (i.e. their loadings; Hopke 1992). The decision of selecting the number of factors in the analysis was based on the eigenvalue for the factors. For this study, factors having an eigenvalue greater than 1 were retained in the model.

After the identification of the probable source of a factor (based on the elements with high factor loadings; Tanner and Wong 2000), its quantitative contribution to total measured PM was described according to a multilinear regression analysis (MLRA; Harrison et al. 1996). The final products of the analysis were equations that predict the contribution of each source for each variable in every sampled day.

2.3 Atmospheric Transport Scenarios

Five day air mass isentropic back-trajectories supported by the HYSPLIT-4 Model (Draxler and Hess 2004) were run at 750, 1,500 and 2,500 m a.s.l. in Cartagena. This tool identified the geographical characterization of air masses as one of the eight sectors appearing in Fig. 2. Most of the time (36% of days in 2004), Cartagena is under Atlantic conditions (including North Atlantic, Northwest Atlantic and West Atlantic air masses) which do not produce remarkably high level episodes for PM10 or PM2.5 sizes. On the other hand, Southwest Atlantic air masses (4% of days in 2004) usually present marked increased levels, probably because of the contribution of African dust. PM levels are the highest with air masses arriving directly from Africa (29% of days in 2004) implying high levels of crustal aerosols. When micro and mesoscale conditions predominate (15% of days in 2004), apart from the intensification of secondary processes (Querol et al. 2004a), there is a stagnation which leads to resuspension of regional material (Gangoiti et al. 2001). It is also possible that sulphate enriched masses reach Spain during the coolest months (14% of days in 2004), and these trajectories are frequently associated with a European origin (Querol et al. 2000). Finally, if the air mass passing over Cartagena has a Mediterranean origin (2% of days in 2004), the resulting meteorological instability often ends in precipitation if the squalls are centred on the Gulf of Cadiz (Capel Molina 1986).

Fig. 2
figure 2

Defined sectors for the air masses origin. AN Northern Atlantic, ANW Northwestern Atlantic, AW Western Atlantic, ASW Southwestern Atlantic, NAF Northern African, ME Mediterranean, EU European, RE Regional

Full size image

Additional information describing these scenarios can be obtained from:

  1. 1.

    Dust Regional Atmospheric Modelling (DREAM; Nickovic et al. 2001), which offers a simulation of the advections of the African dust aerosol at a surface level.

  2. 2.

    SKIRON Forecast System (Kallos et al. 1997), which simulates the vertical integrated dust load.

  3. 3.

    Maps of total optical depth for sulphate, dust, and smoke; as well as maps of sulphate mass, dust and smoke (NRL 2003).

  4. 4.

    SeaWIFS-NASA Project (McClain et al. 1998), which offers satellite images.

The conjunction of the cited maps together with the chemical analysis of the PM filters builds the background for the full characterisation of pollution episodes in Cartagena.

3 Results and Discussion

3.1 PM Levels

Tables 1 and 2 show the levels of PM measured at the two stations. The mean PM10 level at Bastarreche was 47 μg m−3 for the 98 days sampled with 45 of the days registered levels above 50 μg m−3. On the other hand, the mean PM2.5 level at Santa Ana was 22 μg m−3 suggesting that Santa Ana is in the usual range for Spanish urban background stations (Querol et al. 2004b). Only ten of the days sampled registered PM2.5 greater than 35 μg m−3. The thresholds discussed here are given by the Council Directive 1999/30/EC of 22 April 1999 and the recommendations from CAFE (2004).

Table 1 Composition of PM10 at Bastarreche station
Full size table
Table 2 Composition of PM2.5 at Santa Ana station
Full size table

The seasonality of PM levels, as well as the concentration of chemical species through the year, is the consequence of the combination of meteorological history, emission sources and African dust contributions. For example, Fig. 3 shows the concordance between a number of PM peaks and African Dust Outbreaks (ADO). This is consistent with the observation that 36% of days having PM10 levels >50 μg m−3 coincided with African trajectories. As expected (Fig. 4a), under these conditions, the mineral portion is highest (20 μg PM10 m−3, 32%), followed by the Secondary Inorganic Compounds (SIC) fraction (13 μg PM10 m−3, 21%). Because this is a traffic oriented station, OM+EC (11 μg PM10 m−3, 18%) is the third major fraction. This increase of SIC has previously been observed in the Canary Islands during dust outbreaks and could be the result of the interaction of gas species like NO x , HNO3, SO2 or H2SO4 with the African mineral dust, particularly CaCO3, leading to accelerated gas to particulate conversion (Alastuey et al. 2005). The remaining 64% of days with PM10 levels >50 μg m−3 (Fig. 4b) are associated with local/regional sources including traffic as OM+EC (15 μg PM10 m−3, 26%), followed by SIC (11 μg PM10 m−3, 19%) and the mineral fraction (11 μg PM10 m−3, 19%). This description would be consistent with motor vehicle exhaust, the abrasion of tires and breaks, and the resuspension of crustal materials on road pavement (Sternbeck et al. 2002; Wåhlin et al. 2006). As previously established, the mineral fraction in days without ADO could also be increased because of the resuspension from road pavement as well as the resuspension of mineral materials on a regional-scale from non-vehicular sources.

Fig. 3
figure 3

Sequence of the mean daily levels of PM10 at Bastarreche, PM2.5 at Santa Ana and those days when ADO were detected. The straight line points out that when more than 60 μg PM10 m−3 were detected, there was coincidence with ADO

Full size image
Fig. 4
figure 4

Distribution of PM components: a when PM10 levels are above 50 μg m−3, during ADO; b when PM10 levels are above 50 μg m−3, when African dust was not present; c mean composition of PM2.5 during those days with higher levels than 35 μg m−3; d contribution of each identified source for PM10 at Bastarreche; e contribution of each identified source for PM2.5 at Santa Ana. Undet. Non determined fraction; Zn the local Zn metallurgical industry; Industrial-Cartagena a blend of the industrial activity found in relationship with the secondary aerosol

Full size image

3.2 Chemical Composition

The chemical fractions were defined as mineral, SIC, OM+EC, marine aerosol and trace components. On average, the non-allocated portion varied from 29% PM10 at Bastarreche to 34% PM2.5 at Santa Ana station, somehow slightly higher values to those described elsewhere (Querol et al. 2001b; Viana et al. 2002). This could be attributed to the increased relative humidity during the sampling and the presence of organic components from natural sources, such as amino-acids or humid substances detected in coastal rainwaters (Kieber et al. 2005), or from local anthropogenic sources including petrochemical and plastic industries. Concerning the humidity, a negative correlation between the marine fraction and the determined mass was identified. The proposed interpretation is that because of the Mediterranean location of Cartagena, when marine aerosol loaded air masses arrive in the city, they also carry elevated water content, and the percentage of mass determined decreases. The results presented in Tables 1 and 2 seem to agree with what could be expected from a traffic-oriented station, as the highest PM10 component is OM+EC (11.3 μg m−3, 24%). On the other hand, the major component in PM2.5 is the SIC fraction (7.0 μg m−3, 31%). For the ten days when the concentrations of PM2.5 aerosols were higher than 35 μg m−3 (Fig. 4c), there was a high non-determined fraction (16.3 μg m−3, 40%), followed by SIC (12.3 μg m−3, 30%) and OM+EC (6.3 μg m−3, 15%). The levels of SIC during these days are noteworthy, and might be explained both by the recirculation of air masses in the Mediterranean basin (Querol et al. 2000; Millán et al. 2002) and by the oxidation of gaseous precursors emitted at a regional scale (Andreae et al. 2003; Perraudin et al. 2005).

3.3 PM Sources: Bastarreche and Santa Ana

Once FA was applied, six and five sources were determined at Bastarreche and Santa Ana, stations respectively. Tables 3 and 4 show the factor loadings of the principal species weighing in the factors (i.e. the sources) and their percentage of variance explained by each factor. At the two stations, the mineral source explains the major variance; this source includes the natural minerals from ADO, the local/regional resuspension (Eltayeb et al. 2001; Querol et al. 2001a; Alastuey et al. 2005) and anthropogenic emissions coming from demolition and building. The traffic source is common at both of the stations as the second source that explains the largest fraction of the variance of the data. As expected, the highest component loadings in this factor are for OM+EC, Sb, \({\text{NO}}_3^ - \). The marine source represents 5% and 7% of variance at Bastarreche and Santa Ana respectively, with Cl and Na as the main tracers in this factor. A factor interpreted as the local Zn metallurgy industry accounts for the 7% and 5% of variance at Bastarreche and Santa Ana respectively, with Zn, Cd and Pb as its markers, consistent with mineralogy and processing of spharelite (del Valle Alonso et al. 1970). The secondary inorganic aerosols at Bastarreche account for 8% of variance, with \({\text{SO}}_4^{2 - } \), \({\text{NO}}_3^ - \), \({\text{NH}}_4^ + \) as main tracers. A mixture of compounds appear clustered at Santa Ana, which might represent aged air masses containing the emissions from the local industries next to the city centre (\({\text{SO}}_4^{2 - } \), V, Ni, As, \({\text{NH}}_4^ + \)). Finally, the last factor at Bastarreche is the one identified as the military shipyard (5% of variance) with Cr and Ni as main tracers.

Table 3 Results of the PCA for the PM10 fraction at Bastarreche
Full size table
Table 4 Results of the PCA for the PM2.5 fraction at Santa Ana
Full size table

Once the sources have been identified, their daily contribution to the total aerosol mass was calculated using MLRA. Based on this analysis, Fig. 4d and e show each mass contribution of individual sources to the total levels of PM.

In summary, the association among sources and the chemicals that they emit are presented in Tables 5 and 6: There is general agreement between the type of source and their expected composition (Canadian Chemical Producers’ Association 2001). Two specific sources in Cartagena are the Zn metallurgy and the military shipyard. The first one is characterised by Zn in the smallest size fraction and Cd, Zn and Pb in the coarsest size. The emissions from the shipyard are only detected in the nearest station to the harbour (i.e. Bastarreche). The shipyard slightly contributes to Cd, Sr and P although, as established in Table 5, its main tracers were Cr and Ni (U.S. Department of Navy 1996).

Table 5 Percentage contribution from each identified factor to the chemical species analysed in PM10
Full size table
Table 6 Percentage contribution from every identified factor to the chemical species analysed in PM2.5
Full size table

3.4 Characterization of ADO

The identification of ADO was performed according to Querol et al. (2004a). However, no different factor for these ADO was distinguished from the mineral fraction in PCA (Tables 3 and 4). To better establish if the chemical species are carried in the African dust or they are indigenous in Cartagena, we propose the ratio of each chemical species divided by the total mass of the crustal compounds in days under the influence of ADO and days with no ADO (Table 7). In this way, those chemical elements (e.g. Al2O3, Ti, Rb) whose mean of the ratios in days with ADO was statistically higher (level of significance < 0.001) than mean of the ratios in days without ADO could be used as tracers of the influence of the African events. Conversely, those chemical elements (e.g. Ca, Cu, Pb) whose mean of the ratios in days with ADO was statistically lower (level of significance < 0.001) than mean of the ratios in days without ADO could be used as tracers of the local aerosol influence in Cartagena: in other words, in ADO days those elements with the mentioned decreased ratio would experiment a dilution by the African dust. The results presented in Table 7 are in agreement with the bibliography for African tracers (Salvador et al. 2004; Viana et al. 2006) and the elemental composition in local soils [i.e. Cu and Pb occur notoriously in local soils (Moreno-Grau et al. 2002)].

Table 7 Significant differences (level of significance < 0,001) in the ratio “mass of each chemical species divided by the total mass of mineral fraction” found in days with ADO
Full size table

4 Conclusions

Chemical composition of PM10 and PM2.5 was determined at a traffic hotspot and a suburban location in the Southeast of the Iberian Peninsula. A Factor Analysis allowed us to observe the following PM sources: crustal, traffic, secondary, marine, Zn metallurgy industry and shipyard. The crustal source includes local dust resuspension, ADO and road dust resuspension. The secondary aerosols source represents aged air masses (i.e. \({\text{NH}}_{\text{4}}^{\text{ + }} \), \({\text{SO}}_4^{2 - } \), \({\text{NO}}_3^ - \)) together with the contributions from the power station (V, Ni). A Zn metallurgical industry source was described by Cd, Zn and Pb in the two grain sizes. Finally, a source from a shipyard had a minimal contribution, was detected only in the coarser fraction, and was described by Cr and Ni. In the cases when the city is under environmental conditions which increase the PM levels because of natural causes, the clear definition of the reasons (i.e. anthropogenic or natural) of high PM levels were identified based on isentropic back-trajectory air masses and aerosol maps.