Saharan dust levels in Greece and received inhalation doses

The desert of Sahara is one of the major sources of mineral dust on Earth, producing around 2 × 10 8 tons/yr. Under certain weather conditions, dust particles from Saharan desert get transported over the Mediterranean Sea and most of Europe. The limiting values set by the directive EC/30/1999 of European Union can easily be exceeded by 5 the transport of desert dust particles in all south European areas and especially urban. In this study, the e ﬀ ects of dust transport on air quality in several Greek urban areas are quantiﬁed. PM 10 concentration values from stationary monitoring stations are compared to dust concentrations for the 4-year period 2003–2006. The dust concentration values in the Greek areas were estimated by the SKIRON modelling system 10 coupled with embedded algorithms describing the dust cycle. The mean annual dust contribution to daily-averaged PM 10 concentration values was found to be around or even greater than 10% in the urban areas throughout the years examined. Natural dust transport may contribute by much more than 20% to the annual number of exceedances – PM 10 values greater than EU limits – depending on the speciﬁc moni- 15 toring location. In a second stage of the study, the inhaled lung dose received by the residents in various Greek locations is calculated. The particle deposition e ﬃ ciency of mineral dust at the di ﬀ erent parts of the human respiratory tract is determined by applying a lung dosimetry numerical model, which incorporates inhalation dynamics and aerosol physical processes. The inhalation dose from mineral dust particles was 20 greater in the upper respiratory system (extrathoracic region) and less signiﬁcant in the lungs, especially in the sensitive alveolar region. However, in cases of dust episodes, the amounts of mineral dust deposited along the human lung are comparable to those received during exposure in heavily polluted urban or smoking areas.

coupled with embedded algorithms describing the dust cycle. The mean annual dust contribution to daily-averaged PM 10 concentration values was found to be around or even greater than 10% in the urban areas throughout the years examined. Natural dust transport may contribute by much more than 20% to the annual number of exceedances -PM 10 values greater than EU limits -depending on the specific moni-15 toring location. In a second stage of the study, the inhaled lung dose received by the residents in various Greek locations is calculated. The particle deposition efficiency of mineral dust at the different parts of the human respiratory tract is determined by applying a lung dosimetry numerical model, which incorporates inhalation dynamics and aerosol physical processes. The inhalation dose from mineral dust particles was

Introduction
Mineral dust, produced by wind erosion, over arid and semi-arid areas of North Africa, get transported over the Mediterranean Sea and most of Europe, under certain weather conditions. Desert dust particles affect the atmosphere and climate. By scattering and absorbing solar radiation they modify the planetary albedo and reduce the amount of 5 radiation reaching the Earth's surface (Yu et al., 2001). Aerosols act as cloud condensation nuclei (CCN), thus modifying the microphysical, micro-chemical and, hence, optical and radiative properties of clouds (Charlson et al., 1991). They can also influence the nutrient dynamics and biogeochemical cycling of both terrestrial and oceanic ecosystems (Herut et al., 2001). 10 In the past decades, a clear connection between suspended particulate matter and health effects has been found (Dockery and Pope, 1994). Therefore, the European Community has established the Air Quality Directive 1999/30/EC limiting the concentration levels of PM 10 ; mean annual PM 10 concentrations should not exceed 40 µg m −3 , and the daily limit value of 50 µg m −3 should not be exceeded more than 35 days/year.

15
The above standards should have been accomplished by the year 2005. These limiting values can easily be exceeded by naturally produced aerosols (desert dust) for many days during the year in all southern European areas and especially urban. Therefore, recent studies have focused on the estimation of the influence of the African dust on air quality in Mediterranean urban areas (Rodríguez et al., 2003;Escudero et al., 2007). 20 In a recent study in Greece, it was found that for about 50% of the daily exceedances observed in ambient air monitoring stations there was a contribution from Saharan dust transport (Kallos et al., 2007a). The first objective of this study is to quantify the effects of dust transport on air quality in several Greek urban areas using a large number of data from PM monitoring stations 25 and air quality model outputs. The model used is the SKIRON weather forecasting system with the desert dust module (Kallos et al., 1997;Nickovic et al., 2001). The second objective is to investigate the inhaled doses of dust received by the residents in the Greek urban areas during various exposure conditions. Inhaled dust particles may be deposited at different parts of the human respiratory tract (nasopharyngeal, tracheobronchial, alveolar region). The determination of the particle deposition at different parts of the respiratory system is an essential step in investigating possible risks from exposure to desert dust. To this end, a recently developed dosimetry model (Mitsakou 5 et al., 2005) is applied. The model calculates deposited inhaled aerosols in the various regions of the human respiratory tract, taking into account the aerosol properties (e.g. size distribution, hygroscopicity) and the physiological conditions of the individual.
The methodology adopted for the air quality and dosimetry analyses is described in Sect. 2, while Sect. 3 provides discussion of the results. Finally, the concluding remarks 10 of the study are summarized in Sect. 4.

Air quality analysis
For the purpose of this study, the operational runs of the SKIRON system and the dailyaveraged PM 10 mass concentration measurements from air quality monitoring stations 15 of the Greek Ministry of Environment City Planning and Public Works (MECPPW) have been analyzed. Correlations between the model outputs and the PM 10 measurements from the monitoring stations can help identifying the dust contribution to the air quality in the Greek urban areas. The model predictions and the experimental measurements cover the four-year time period [2003][2004][2005][2006]. A brief description of the implemented 20 tools is given below.

SKIRON modelling system
In this study, the SKIRON atmospheric modelling system was used (Kallos et al., 1997;Papadopoulos et al., 2002). SKIRON is a modified version of the Eta/NCEP regional Introduction atmospheric model, fully coupled with a module for describing the desert dust cycle in the atmosphere. The Eta model is well-documented and detailed descriptions of its dynamics and physics components can be found in several studies (e.g. Mesinger et al., 1988;Janjic, 1994, and references therein). This initial version has been further improved by incorporating high resolution sea surface temperature SST fields, spa-5 tial distributions of clay amounts, non-hydrostatic corrections, as suggested in Janjic et al. (2001), in-cloud deposition efficiency, etc. During the SKIRON and POSEIDON projects, the atmospheric radiation and surface processes have been tested and updated (Papadopoulos et al., 2002). The first dust cycle module was developed at the University of Athens in the frame-10 work of the MEDUSE project and further tested and developed during the ADIOS project. In Nickovic et al. (2001), it was given the name DREAM (Dust REgional Atmospheric Model). The present version incorporates several state-of-the-art parameterizations for the description of the production, transport and removal processes of the desert dust cycle, while several inconsistencies in the initial version described in 15 Nickovic et al. (2001) have been corrected. For the definition of the areas that are potentially active, global datasets are used for the vegetation (Olson World Ecosystems with 30" × 30" resolution) and soil texture (FAO/UNESCO with 2 ′ × 2 ′ resolution). The soil texture dataset has been enhanced by using the concept of rocky soil. Also, the soil moisture and turbulent state of the atmosphere play a critical role in the amount of dust 20 uplifted and injected into the air. Since the dust cycle module is dynamically coupled to the atmospheric model, the prognostic atmospheric and hydrological conditions are used to calculate the effective rates of the injected dust at each time step.
The initial version was using a single particle size (1997)(1998)(1999)(2000)(2001)(2002)(2003). The model version used in the present study had four size bins with centered diameters of 1.5, 12, 36, and 25 76 µm (Tegen and Fung, 1994). For the present analysis, the model outputs that correspond to the two smaller-diameter (1.5 and 12 µm) size bins are taken into account in order to approach the PM 10 African dust. The ability of the model to predict the dust concentration for a long period has been tested for long-range transport of Saharan Introduction the Internet site http://forecast.uoa.gr/, while the systematic backup is executed since the beginning of 2000. After 2003, the four-size distribution version with the corrections mentioned above is set up in operational use. Recently, the operational version has been replaced by a newest one with eight size bins, as proposed by Zender et al. (2003), together with several other improvements. In the present work, this long 15 time series of desert dust data with four size bins is used to quantify the effects of naturally produced aerosols in PM 10 monitoring stations in Greece.

Experimental data
The MECPPW has a relatively dense monitoring network in Athens and in major Greek urban areas. However, in many of them there are long gaps in monitoring. Thus, four 20 monitoring sites were selected for the present analysis based on the data availability and geographical location. The selection criteria were related to the length of the time series, the location of the stations and the representativeness of the greater urban conglomerate. Each monitoring station is situated in one of the most populated Greek urban areas, namely, Athens, Patra, Volos, Heraklion. The locations of the monitoring 25 stations are displayed in Fig. 1. According to the criteria proposed by the European Environment Agency, the monitoring station of Athens located northwest of the city center (Thrakomakedones) is a suburban-background station; thus, the contribution of 11972 Introduction

Back Close
Full Screen / Esc

Printer-friendly Version
Interactive Discussion local PM sources (traffic, construction works and other anthropogenic activities) to air quality at the specific area is limited and the contribution of the natural sources can be detected more easily. However, no suburban monitoring stations have been placed in the other urban areas (Patra, Volos, Heraklion), but these cities are relatively small (150 000-300 000 inhabitants). Therefore, the collected data from the urban monitoring 5 stations of the three areas were analyzed. The concentration of PM 10 aerosol particles is monitored at the above-mentioned sites by using β-attenuation systems.

SKIRON output processing
The forecast model generates hourly dust concentrations. Thus, the daily-averaged values of dust concentration had to be calculated based on the hourly values for the 10 comparison with the experimental ones from the monitoring stations. The horizontal grid increment of the model domain is 0.24 • (approximately 25 km). In the vertical 32 levels were used stretching from the ground to the model top (20 km). The determination of the simulated dust concentration at the specific measuring points of the MECPPW stations was achieved by applying linear interpolation to the model outputs.

15
More specifically, the aerosol concentration at the measuring points was approximated by the relative contribution of the closest grid points of SKIRON domain introducing a weight function (inverse distance weighting) for the contribution of the grid points to the actual location of the MECPPW station. 20 For the dosimetric calculations a mechanistic inhalation dosimetry model is used, which includes both inhalation and aerosol dynamics (Mitsakou et al., 2005). The model solves the aerosol general dynamic equation (GDE) along the flow direction (in one dimension). The size distribution is described with a sectional representation, which allows for arbitrary functional forms of the size distribution. The description of the res-ACPD 8,2008 Saharan dust levels in Greece and received inhalation doses rating time-varying alveolar region during breathing (expanding and contracting rhythmically during respiration). The particle deposition is assumed to be the result of the mechanisms of gravitational settling, Brownian diffusion and inertial impaction, acting simultaneously. The air velocity in the airways is determined by solving the equation of continuity.

5
The deposition rate of inhaled particles in the respiratory tract is calculated by summing up the calculated rate for each size bin, taking into account the tidal volume (volume of air inspired and expired during each respiratory cycle) and the breathing frequency (number of breaths per time), as follows where DF is the calculated deposited fraction, n i the particle number concentration of the i th size section at the entrance of the respiratory tract, V T the tidal volume and f the breathing frequency. The physiological parameters V T and f depend on the level of physical exertion. The values utilized in our calculations are taken from ICRP (1994), as recommended for an adult Caucasian male under the typified conditions of light 15 exercise (V T =1250 ml, f =20 min −1 ).

Results and discussions
3.1 Natural dust transport A first approach for the origin of the urban particulate matter is attempted via a statistical analysis of the PM 10 concentration values from the monitoring stations. The ACPD 8,2008 Saharan dust levels in Greece and received inhalation doses Peak PM 10 values can be identified mostly during the transition seasons (spring and autumn) and secondarily in winter in almost all monitoring sites, while minimum values are usually noticed during the summer period due to prevailing northerly wind patterns (etesian winds, see Kallos et al., 1993). For example, the monthly averaged PM 10 concentration for the Athens suburban station is maximum in April, when African dust 5 concentration is also high (see Figs. 2, 3). The most significant sources of PM 10 could be natural (Saharan dust, sea salt, pollen from pine and olive trees during spring), but also anthropogenic (central heating, traffic, industry), as mentioned in previous studies (e.g. Querol et al., 2004). The maximum sea salt concentration occurs during the winter period due to the high sea spray production (Koçak et al., 2007). Moreover, the use of 10 central heating is maximum in winter. Increased dust transport takes place in the transition seasons (spring and autumn, Gerasopoulos et al., 2006;Kallos et al., 2007b). According to the above analysis, PM 10 maximum values in spring or autumn months could be attributed mainly to mineral dust transport. Thus, the analysis based on the measured PM 10 values indicates a noticeable contribution of mineral dust to urban air 15 quality. On the other hand, PM 10 values in a town of approximately 150 000 inhabitants in central Greece (Volos) show maximum aerosol concentrations in the winter period (December). Volos monitoring site is the more northern site comparing to the other examined sites, the distance from Heraklion is approximately 4 • (∼400 km). The influence of dust transport decreases towards higher geographical latitudes. The effect of 20 dust transport is reduced with increasing distance from the dust source, as dust particles are deposited via wet and dry mechanisms. Furthermore, there are two extended industrial areas at the outskirts of the town, where high PM amounts are released. The first industrial area is northwest of the town and the second one (concrete production) is at the southeast part. 25 The time periods of increased dust transport are also pointed by the simulated dust concentration values. In Fig. 3 Model estimates of near ground dust concentrations have been compared with PM 10 observations. More specifically, in Fig. 4a-d scatter plots between predicted dust concentrations and measured PM 10 concentrations for the cases of daily exceedances (observed PM 10 values greater than the daily EU limit of 50 µg m −3 ) are presented for each monitoring station. The trend lines have been derived for each dataset and the 10 correlation coefficients have been calculated. The seasonal trend lines of the scatter plots and the correlation coefficients for each season of the year separately and for each monitoring site are presented in Table 1, where x is the observed PM 10 concentration and y is the simulated dust concentration (for cases of daily exceedances). The correlation coefficients appear to fluctuate between 0.78 (Heraklion, autumn period) 15 and 0.26 (Volos). The low coefficient value between PM 10 and dust concentrations found for Volos station during all seasons of the year limits the statistical confidence of the estimation of dust contribution to the urban air quality for the specific area. This low correlation could be attributed to the geomorphological characteristics of the area, more specifically, the town is adjacent to a steep mountain that receives considerable 20 amounts of precipitation and, thus, the washout effect becomes important. On the other hand, higher correlation values have been calculated for the southern town of Heraklion, especially during the transition autumn period. In general, the correlation between measured and simulated particle concentrations appears significantly weaker during the summer period in all sites, when dust transport from Saharan desert is limited due 25 to the northerly etesian winds.
The slope of the trend line in Fig. 4a-d

Printer-friendly Version
Interactive Discussion slope of the trend line is 0.39 (Fig. 4a) that denotes a dust contribution to urban aerosol approaching 40%. Gerasopoulos et al. (2006) performed a chemical analysis of PM 10 in combination with back-trajectories analysis for Heraklion during dust episodes. The analyses showed a participation of mineral dust at about 40% in PM 10 daily exceedances throughout the year. Thus, the dust contribution calculated in the 5 present analysis seems to be verified by the analysis of Gerasopoulos et al. (2006). The fraction of dust transported particles to PM 10 exhibits the lowest value during the summer period (the slope of the trend line is 0.19), which is in agreement with the findings of the latter work. The high intercept values for the site of Patra demonstrates the remarkable influence of local pollution, while the slope of 0.76 during the spring 10 period reveals that dust transport also plays a significant role in the urban air quality. Dust transport seems to affect PM 10 values in cases of daily exceedances in Athens by 25 and 34% during the spring and autumn periods respectively. However, a comparison with the other seasonal periods is not feasible due to very limited number of PM 10 observations during the winter and summer periods. It should be noted, though, that air quality in the Greek urban areas examined here is influenced by the introduction of sea-salt particles, as all the studied areas are seaside. Moreover, southwestern flows that enhance dust transport from the African continent, could also lead to severe air pollution episodes as they are combined with an abrupt temperature change (Kallos et al., 1993). Therefore, PM 10 concentration data could not distinguish the dust trans-20 ported mass from the other suspended materials. Note that the different correlations between simulated and observed values for the various urban areas should be also attributed to the different PM sources. More accurate predictions of dust contribution are obtained for the monitoring site of Athens suburb, as the monitoring station there is characterized as suburban-background. 25 A more explicit statistical analysis of the observed and simulated data for the four years that we examined has been performed. Some of these results are summarized in Table 2. In this table, the averaged observed and simulated values, the dust contribution to averaged PM 10 concentrations, the number of annual exceedances and the ACPD 8,2008 Saharan dust levels in Greece and received inhalation doses period can be attributed to a possible reduction of dust episodes, since mineral dust transport performs a significant inter-annual variation (e.g. Sunnu et al., 2008), while no particular reduction in anthropogenic PM emissions in Greece has been reported during the period analyzed. The dust contribution to PM 10 concentrations -annual average of daily ratios of dust to PM 10 concentrations -approaches and exceeds 10% 20 for the areas of Patra, Athens suburb and Heraklion. In terms of daily exceedances the dust contribution is very high at the suburban station of Athens (e.g. 54.2% for 2005) and noticeable for the other Greek urban areas (>20%).

Inhaled dose
The human exposure to Saharan dust particles under various dust concentration levels   -90, 90-95, 95-9.5, 99.5-100). For each exposure category, the dust mass concentration is assumed to take the maximum value. By that approach, the upper limits of human exposure and received lung dose to Saharan 5 dust particles can be derived. In Table 3, the number of days that correspond to each exposure category is presented (the boundaries of each exposure level, as defined by the percentiles, are shown in the parentheses). As expected, the percentiles of dust concentration exhibit their highest values at Heraklion, while the lowest percentiles are noticed for the town of Volos. 10 Based on the different exposure categories, we assess the lung dose for residents of the Greek urban sites due to Saharan dust transport throughout a year. However, the four-size SKIRON version used in the present analysis could not provide a detailed particle size distribution, as required by the dosimetry model. Thus, for the description of the size distribution of the dust transported cloud in the Greek area, we employed 15 the long-range transport mode proposed by Schulz et al. (1998) and tested by Zender et al. (2003). More specifically, the dust particle size distribution follows a lognormal form with mass median diameter (MMD) equal to 2.524 µm and geometric standard deviation equal to σ=2. For the mineral aerosol, the particle density ρ p is taken equal to 2.65 g cm −3 . Note that, a lognormal distribution is fully described by the parameters 20 MMD, σ and mass concentration m. The long-range transport mode that was used for the present calculations appears to coincide with the "coarse 2" mode derived from the analysis of Gerasopoulos et al. (2007). The latter authors studied the seasonal mass size distributions at a coastal site situated 70 km east-northeast of Heraklion (Finokalia). For the spring period, enhanced concentration for particles with aerody-25 namic diameters 3-7 µm (physical diameters 1.8-4.3 µm) was obtained, attributed to dust transport from Northern Africa. The normalized particle mass distribution (particle mass distribution divided to mass concentration) with the physical characteristics (MMD, σ) used for the dosimetric calculations are illustrated in Fig. 5, as a  Printer-friendly Version Interactive Discussion particle diameter D. However, it should be noted that long-range transported particle sizes measured away from source regions may depend significantly on the distance from the source regions (Park and Kim, 2006). Larger particles can be detected at regions near the source points. The mode of the particle size distribution appears shifted towards smaller particle diameters, as the distance from the dust sources increases.

5
Therefore, the aerosol size distribution used for the depiction of suspended particles at the more distant sites (Athens suburb, Patra, Volos) may be distorted. Aerosol deposition at the various regions of the respiratory tract is strongly dependent on particle size distribution of the inhaled aerosols (Lippmann et al., 1980). Thus, the incorporation of a more detailed bin method in SKIRON modelling system would lead to more 10 accurate dosimetric calculations. However, the use of the current model is considered as efficient for a generic dose assessment. In SKIRON modelling system, the atmospheric dust behaves as an inert chemical substance. Therefore, in the present study, mineral dust is considered as nonhygroscopic as is at its native state. However, dust particles may become hygroscopic 15 through the "aging" process -chemical alterations at the particle surface (Zhang et al., 1994;Levin et al., 1996). Moreover, the presence of hygroscopic biological aerosols (i.e. bacteria, pollen) in the atmosphere (Möhler et al., 2007) is often linked to dust transport (Shinn et al., 2003). Particle deposition pattern along the human lung is highly dependent on aerosol hygroscopicity; thus, the forecast model would be sig-20 nificantly improved by taking into account the physicochemical properties of the dust particles in a forthcoming study. Nevertheless, the assumption of inert dust particles, as commonly adopted by the models that deal with dust transport (e.g. Nickovic et al., 2001, and references therein), is applied for the current dosimetric calculations.
The ratio of the deposited mass over the total inhaled mass is presented in Table 4 25 for the different regions of the human respiratory tract (extrathoracic -ET, tracheobronchial -TB, alveolar -AI) and for one breathing cycle. The calculated values of the deposited fraction show that most of the PM mass stays in the upper respiratory system, namely the ET region. The deeper lung regions (AI) have a lower concentra- Interactive Discussion tion of the larger inhaled particles as a result of the deposition made at the entrance of the respiratory tract (ET region). The great amount of PM mass is accumulated at the larger particle sizes, thus the airborne particle mass decreases while traversing the lung. As a consequence, only 10% of the inhaled particle mass is deposited in the pulmonary lung (AI region).

5
In order to obtain an integrated assessment of the received inhaled dose from Saharan dust to the residents of the Greek urban areas, the daily mass deposition rates in the lung (TB and AI regions) were calculated for the different exposure conditions throughout a year (Fig. 6), assuming a typical daily exposure of 6 h. Extremely high mass deposited values during a day are performed for the days of severe dust episodes 10 (99.5th-100th percentile), reaching up to 600 µg day −1 for a resident at Heraklion. On the other hand, the dust mass deposited in the lung for a resident at Volos is only half under all exposure conditions examined here. It must be noted, that the relationship between inhaled dose and dust concentration is linear, since identical physical activity (light exercise) was assumed for all exposure conditions and at the various sites (see 15 Eq. 1). The calculated mass deposition rates are compared to the respective deposition rates presented in the work of Venkataraman and Raymond (1998), so as to interpret the present results. The latter authors reported that the lung dose resulting from the typical daily exposure in the polluted Los Angeles basin equals to 294 µg day −1 , while 16-h exposure to environmental tobacco smoke (ETS) leads to dose ranging from 135 20 to 187 µg day −1 for adult males and from 76 to 106 µg day −1 for females. Thus, the received lung dose for a resident in Patra or Heraklion town during a moderate dust episode (95th-99.5th percentile) is similar to the dose received after exposure in the polluted urban air basin of Los Angeles. Under low dust exposure conditions (90th-95th percentile), the equivalent doses may approach the particle respiratory deposition 25 in a smoking area.
We have not addressed the toxicological significance of this inhaled dose resulting from the exposure in mineral dust particles, if any. The current calculations showed, however, that the respiratory dose resulting from dust exposure could be significant,

Conclusions
In the present study, the determination of the mineral dust levels in Greek urban areas and the received inhaled dose was attempted. The quantification of the dust transport effects in urban air quality was accomplished by analyzing a large database of PM 10 5 daily measurements and SKIRON model predictions. The analysis confirmed that the dust transport effect is maximized during the transition spring and autumn periods, as has already been noticed in previous studies. The concentration of mineral dust particles decreases from South to North, due to particle deposition mechanisms, thus, the fraction of dust to PM 10 concentration appeared to vary significantly depending on 10 the location. Natural dust transport may contribute by much more than 20% to the annual number of exceedances. The mean annual dust contribution to daily-averaged PM 10 concentrations was often found to be greater than 10% throughout the years examined, while the contribution of mineral dust transport to the annual number of exceedances exceeded 20% in most of the studied areas.

15
The determination of the received inhaled doses by the residents in the Greek urban areas showed that the inhaled dose was greater in the upper respiratory system (extrathoracic region), since the major amount of PM mass is accumulated at the larger particle sizes that are mostly deposited before reaching the lung. Higher lung deposition rates were calculated for the citizens of the southern part of the country (Heraklion), 20 while the doses are reduced by half for an individual living 400 km far to the North. The calculated lung doses during severe and moderate dust episodes were found to be comparable to those received during exposure in polluted urban and smoking areas.
These concluding remarks are also applicable to most South European urban areas, since dust episodes and local weather conditions are similar to the Greek ones. ACPD errors in dust vertical distributions over Rome (Italy): Multiple particle size representation and cloud contributions, J. Geophys. Res., 112, D15205, doi:10.1029/2006JD007427, 2007