The subsequent discussion is structured so as to provide insights on: 1) how and why the amount of dust and its magnetic signal reflects temporal and lateral changes in environmental conditions; ii) which magnetic minerals are responsible for the magnetic signature and what are their grain sizes; iii) what is the relationship between dust magnetism and the content of major heavy metals and PAHs which would justify the utilization of magnetic parameters in health studies; iv) review of physical background, epidemiological and toxicological studies on the role of mineral dust and especially iron oxides for human health; and v) testing the validity and reliability of the magnetic ratio χindoor/χoutdoor as a tool for estimation of cardiovascular diseases caused by urban pollution.
4.1. Site-Specific and Temporal Variations of Mass-Specific Magnetic Susceptibility and Dust Loading Rates
 Mass-specific magnetic susceptibility (χ) as one basic magnetic characteristic of the solid matter, depends generally on the mineralogy, concentration and grain size of strongly magnetic Fe-oxides in the material [Dunlop and Ozdemir, 1997]. In this context, settled dust in urban areas has a magnetic signature, which is determined by the cumulative effect of the magnetic fractions, inherent to its various components. Temporal and lateral changes of the magnetic susceptibility at a given location thus, are indicative about any changes in the relative proportion of the different fractions, or changes in the dust sources. Magnetic measurements are capable of identifying even subtle changes in the content of strongly magnetic Fe-oxides down to 50 ppm (titano)magnetite [Dearing, 1999]. Thus, monthly variations of χ shown in Figure 2 represent a sensitive record of changing mineral content in the outdoor and indoor settled dusts. Systematically lower indoor values suggest decreases in the relative proportion of strongly magnetic iron oxides in the total mineral dust. Gravitational settling is the major process of coarse dust dynamics, forcing sedimentation of coarser and heavier particles close to the origin [Fisher and Macqueen, 1981]. The amount of outdoor Fe-oxide particles, which belong to the heavy mineral fraction, will depend on their size distribution. The observed systematically lower indoor susceptibilities as compared to their outdoor counterparts inFigure 2, suggests that the major source of indoor dust is the penetration of outdoor dust inside by means of diffusion, turbulent flows through open windows (doors) and mechanically introduced dust by people's activities. The absence of correlation between monthly mass-specific susceptibility and PM10 concentrations for most of the sites implies that the main contribution to the magnetic signature of the settled dust is coming from traffic and re-suspended road dust, as far as seasonality in PM10 is generally explained by higher emissions from combustion related to heat production (domestic burning, power plants, etc.). There are different cases, reported in the literature, some of them showing clear seasonal trends in PM and its magnetic susceptibility [Muxworthy et al., 2003; Kim et al., 2007], and others do not show this. For example, no seasonal changes in the coarse PM10 contents are observed by Liu and Harrison in a study of dust from UK, which is explained by the predominantly traffic-related sources. Temporal changes of dust loading rate (Figure 2b) indicate higher sedimentation outdoors than indoors, thus proving our concept that indoor dust sources can be neglected. A particular case is Sofia site S1o and partly SZ1o, which are characterized by lower outdoor dust loading rates as compared to indoor ones. This can be explained by the permanently windy microclimate conditions at these sites, mentioned in Table 1. The range of variation in dust loading rate is well within typical loading rates reported for settled dust [Maertens et al., 2004]. Taking the monthly average loading rates, mean dust load per year is calculated. We obtained values between 6 and 15 g/m2/yr. The highest dust load is calculated for Stara Zagora and especially high value of 71 g/m2/yr for the outdoor site in the city of Pleven - PL2o. As far as we have chosen the entrances of public buildings (mainly schools) for indoor places (Table 1), there should be little or no effects of indoor PM sources on the amount and characteristics of the total settled dust.
 The relative importance of temporal susceptibility changes for each sampled site is represented by the median, lower and upper quartile ranges of χ in Figure 3. The largest variability is observed for the indoor and outdoor dusts from Burgas, and outdoor settled dusts from Pleven and Plovdiv. Using t-test statistics, significant difference atα = 0.05 level between mean χ values of indoor and outdoor sites per each city, is proved. An impression on the absolute differences in magnetic susceptibility of settled dusts from the six cities provides Figure 5. Direct comparison among signals from indoor, outdoor and road dusts implies that except for Burgas, outdoor dust is far much enriched with magnetic fraction as compared to the other two dusts. This finding indicates that fine grained magnetic fraction from anthropogenic activities is settled at higher elevations than the road level but is not fine enough to penetrate (infiltrate) effectively indoors. The exact indoor – outdoor (I/O) ratio would then depend on the particular grain-size distribution of the ambient dust and its aerodynamic behavior. The almost equal values obtained for mass specific magnetic susceptibility of road dust and outdoor dusts from Burgas suggest that in this case coarse-grained soil-derived magnetic particles dominate the total settled dust material.
4.2. Magnetic Mineralogy and Grain Size of the Settled Dust
 Further implications of magnetic signature to environmental pollution need to take into account the type and grain size of the main magnetic minerals contributing to the signal. A suitable tool for this purpose is thermomagnetic analysis of susceptibility (Figure 4). Similar shapes of the heating curves except B2o site imply that the collections of settled dust are characterized by uniform magnetic mineralogy, originating from both local soil dust component and magnetic fraction of the anthropogenic emissions. All curves show the most pronounced decrease of the signal at about 580°C, which is a typical indication of the presence of fine grained magnetite (Fe3O4) [Dunlop and Ozdemir, 1997]. Narrow grain size distribution and relatively large magnetite crystals could be responsible for the sharp decrease of the signal at sample PL2o. The rest of the sites show more gentle convex shapes of the heating curves, suggesting wider grain size distribution of magnetite particles. The decrease of susceptibility after cooling back to room temperature is most probably due to oxidation of magnetite during heating. This process caused also the appearance of hematite component with TN of 700°C for the samples PL2o and SZ1o. Secondary (laboratory induced) origin of hematite is assumed, taking into account the absence of any remanence acquisition in high dc fields during IRM acquisition experiment, shown in Figure 6. The shape of the heating curve for the dust sample from Burgas (B2o) suggests the presence of additional Tc of about 480°C, which can be attributed to titanomagnetite. This titanomagnetite phase is widely found in soils from Burgas region [Jordanova and Jordanova, 2010]. Similar to our data, magnetite as a carrier of the magnetic signal of atmospheric dust is reported in a number of recent studies [Muxworthy et al., 2002, 2003; Sagnotti et al., 2006; Xia et al., 2008; Elzinga et al., 2011]. The grain size of the magnetic fraction in indoor and outdoor dusts could be deduced from the hysteresis parameters, having proven the dominance of magnetite [Dunlop, 1986]. Hysteresis loops closed at fields up to 300 mT (Figure 7a) and the signal saturated are indicative for magnetically soft grains (e.g., magnetite) in accordance with the results from thermomagnetic analyses. Systematically lower remanent coercivities of outdoor dusts relative to indoor pair (Figure 7b) prove the concept that finer particles are settled inside the buildings as a result of gravitational differentiation. Moreover, significantly higher paramagnetic component in outdoor dusts (Figure 7c) points to a higher portion of paramagnetic minerals, deposited outdoors. A possible source of such particles is the soil dust component. A Day plot (Figure 7d) gives further evidence that the effective magnetic grain size of the magnetic fraction is pseudo-single domain. Deviation of the experimental data points from the mixing line of SD + MD grains [Dunlop, 2002] suggests that the magnetic fraction in the dust samples is not a bimodal mixture of coarse and fine grains. Rather, the data set except samples from Burgas (data points from the lower right corner of the PSD region) fits well to the mixing SP + PSD curve (ferrofluid 9.3 nm and 1–3 μm hydrothermal magnetites) from the same work [Dunlop, 2002], revealing possible 10–25% SP component in the settled dust. Comparison with the grain-size dependence of magnetic parameters for magnetite (Mrs/χ and Bcr), compiled by Peters and Dekkers  also implies the predominance of 1–3 μm magnetite grains. Calculated Mrs/χ ratio for 10 indoor dusts gives mean value of 11.33 ± 2.97 × 103 A/m, and value of 9.98 ± 2.58 × 103 A/m for 16 outdoor dust samples, falling in the range of values of Mrs/χ typical for small PSD magnetite grains. This grain size inference is supported by the identified peak at (1.3–1.8))μm in the number concentration of mineral particles (Table 2). Predominance of the fine silt fraction (d < 30 μm) observed under light microscope in the studied samples (Figure 8) implies that urban indoor and outdoor dust contains fine grained particles. Systematically lower proportion of d < 30 μm in outdoor dusts as compared to indoor dusts (Figure 8) proves the general observation that settled dust indoor has smaller mean grain size than the outdoor dust.
 The presence of anthropogenic magnetic particles in the settled dusts is ascertained by the identified spherules (Figures 9a and 9d), which usually originate from combustion processes [McLennan et al., 2000; Kukier et al., 2003]. Smaller Fe-containing clusters of particles (Figure 9b) are most probably exhaust particles from diesel combustion vehicles. Atmospheric dust has also affected the pollen grains, which are characterized by dense cover of fine dust particles on their surfaces (Figure 9c). This adherence of fine particulate matter to the pollen grains is another factor, which increases the allergenic potential of pollens and makes them more aggressive to the immune system [D'Amato et al., 2002, 2007].
4.3. Heavy Metal Content, PAHs and Magnetism
 Contamination of settled dust with hazardous substances is a well-known phenomenon, which is related to emissions from various sources such as vehicles, industrial production, coal and wood firing, etc. [Menichini, 1992; Azimi et al., 2005]. The measured concentrations of the major toxic heavy metals in studied indoor and outdoor dusts (Table 2) indicate significant enhancement of As, Pb, Zn and Cu in all samples, compared with typical values for non-contaminated soils and major rock types [Alloway, 1995]. Based on the results from cluster analysis, data for Burgas (samples B2o and B1o) were considered separately. We examined the possibility of having high HM content inherited from the soil derived and lithogenic component. Previous magnetic and geochemical studies of different soils from Burgas region [Jordanova and Jordanova, 2010] however, report significantly lower concentrations of the listed heavy metals in the soil cover, compared to the composition of the dust samples in the present investigation. Thus, lithogenic component as possible source of heavy metals in Bourgas settled dusts can be discarded. Another possible explanation for the very high regional concentrations of As and heavy metals is identified by moss monitoring of long-distance trans-boundary pollution from Turkey coming from two glass factories [Coskun et al., 2009]. On the other hand, extremely high content of As is accompanied by very high Cu content in the sample B2o from Burgas, suggesting that this pollution may be caused by emissions from Cu-smelting industry in the region.
 Analyses on dust samples have been done for the summer months (June–August 2009), thus avoiding contribution from emissions of heating systems (coal and wood firing domestic and public installations) as one possible source of pollution. According to the analysis of principal components, several factors account for variability in the data set. Factor 1 with the highest loadings by Fe, Mn and χexplain 40.6% of the total variance and can be attributed to soil dust and re-suspended lithogenic component. The second factor, determining another major source of heavy metal pollution by As and Pb could be assigned to vehicle emissions [Johansson et al., 2009]. The third factor with main loadings from Co and Cr may be attributed to industrial activities [Alloway, 1995]. The fourth factor accounts for 9% of the total variance with major contribution from Ni and Cu. The latter element is known to originate largely from brake wear [Hulskotte et al., 2006; Thorpe and Harrison, 2008]. The calculated Tomlinson pollution index (PLI), giving an estimate of the total HM load indicates that most of the samples fall in the category of moderately polluted with PLI in the interval 2–3 [Angulo, 1996]. Plovdiv (P2o) and Burgas (B2o) outdoor dusts can be classified as moderately and highly polluted respectively. The latter two locations are at the closest distance to the roads (see Table 1), suggesting traffic as a main pollution source.
 Tomlinson pollution index (PLI) shows significant at p < 0.05 correlations with Pb, Zn, As and Ni, due to the predominance of these pollutants (Table 4). Mass specific magnetic susceptibility does not correlate significantly with PLI due to two major factors: i) data set is composed of samples from different regions, which have various lithogenic composition of the soil – derived dust component; ii) various pollution sources, whose emissions contain different magnetic fractions. As a consequence, magnetic susceptibility for distinct multiregional data set is not directly related to the degree of pollution.
 Polycyclic aromatic hydrocarbons (PAHs) are formed as a result of incomplete combustion (pyrolysis) or high temperature pyrolytic process during combustion of fossil fuels/organic materials, as well as in natural processes such as carbonization (pyrosynthesis), by-products of incineration of industrial and urban wastes, oil spills and vehicle exhausts. Motor vehicle emissions (especially diesel vehicles) make a considerable contribution to PAH concentration in air due to burning and incomplete combustion of diesel or gasoline. Most of the PAHs are highly toxic substances, playing important role for the human health [Kameda et al., 2005; Chen and Liao, 2006; Srogi, 2007; Maertens et al., 2008]. Although a limited number of samples were analyzed for PAH content (Table 3), it can be concluded that high molecular weight 5-ring PAHs (BbF, BkF, DahA) are the dominant compounds. These are reported as major emissions from diesel vehicles [Marr et al., 1999]. Concentrations of some PAH marker compounds and their ratios can give indication about the impact of different sources of airborne compounds [Guo et al., 2003; Tobiszewski and Namiesnik, 2012]. Table 5 provides the calculated values of diagnostic ratios for PAHs, such as R178 = Ant/(Ant + Flu); R202 = Flu/(Flu + Pyr); R228 = BaA/(BaA + Chr), and R276 = IcdP/(IcdP + BghiP) [Yunker et al., 2002]. Very similar values are obtained for the ratio R202:<0.5 for the samples B2o, R1o, PL2o, S3o and SZ1o, indicating that Ant and Flu originate from combustion in diesel cars; while for the other samples R202 > 0.5, suggesting that these PAHs are produced during vegetation and coal burning [Yunker et al., 2002]. The mean level of ΣPAHs in this study (0.57–8.3 mg/kg) is close to those in Guangzhou, China (0.84–12.3) mg/kg [Wang et al., 2011], Greater Cairo, Egypt (0.045–2.61 mg/kg) and relatively higher than those in the United Kingdom (0.002 mg/kg), Norway (0.0069 mg/kg), Canada (0.0011 mg/kg) and Australia (0.0033 mg/kg) [Hassanien and Abdel-Latif, 2008, and references therein].
 Link between magnetic parameters of urban dust and its PAH content is reported in several studies, which found a linear relationship between PAH content and Mrs [Lehndorff and Schwark, 2004; Halsall et al., 2008; Jordanova et al., 2010]. In the present study, we found dependency between total PAH content and grain-size sensitive magnetic parameters, e.g., coercivity of remanence (Bcr) and the ratio Mrs/χ (Figure 11). It indicates that higher PAH content is linked to dust samples with smaller effective magnetic grain sizes of the magnetic fraction (magnetite). This relationship could be explained in two alternative ways: i) either increased content of PAHs is due to an increase in surface adsorption caused by higher surface/volume ratio with decrease of particle size; or ii) combustion processes responsible for PAH emissions vary in a way that higher amount of PAHs is linked to correspondingly smaller grains of the accompanying magnetic fraction. Affinity of PAHs for sorption onto inorganic particles has been studied by Fang et al. . They show that surface adsorption mechanism is the main sorption process of phenanthrene on zero valent iron, copper and silicon dioxide engineered nanoparticles.
4.4. Magnetic Properties of Settled Dust and Health Indicators
 Extensive number of publications have shown that the concentration dependent magnetic parameters such as magnetic susceptibility (χ) can be used as a proxy for the pollution degree (for reviews see Petrovsky and Elwood  and Evans and Heller ). Only one publication up to our knowledge deals with interdisciplinary research directed toward combined use of magnetic studies and health data. In this work, Morris et al.  found out significant correlation between mutagenicity of the organic compounds of dust from filters, and the values of magnetic susceptibility of the mineral fraction. In this specific example, magnetic susceptibility of dust filters is directly compared to mutagenicity potency, as far as the study area has single pollution source and uniform background magnetic signal from the local lithogenic dust sources.
 General question arise before any attempts to correlate magnetic and health data: what is the direct (or indirect) link between the two parameters? Are magnetic compounds toxic and dangerous for human health? Literature review shows that toxicological and epidemiological studies are reported for nanosized compounds, which are often used in drug delivery as well as for contrasting agents [Weinstein et al., 2010]. Genotoxicity of 1–3 μm magnetite particles and subway particles rich in magnetite was examined by Karlsson et al. , who found much higher DNA damage to human cells by subway particles than that of pure magnetite. Singh et al.  and Mahmoudi et al. reviewed publications on cellular toxicity, DNA damage and oxidative stress caused by superparamagnetic iron oxide nanoparticles, coated or uncoated by various substances used in biomedical applications. Cytotoxicity and genotoxicity of size-fractionated magnetite was studied byKonczol et al. , revealing that the major genotoxicity effect of magnetite particles on human alveolar cells is production of reactive oxygen species (ROS). Furthermore, they demonstrate that magnetite particles induce concentration-dependent DNA damage. A slight size dependency is observed, as the larger particles induced less genotoxicity. Extensive study on size dependence of toxicity of metal oxide particles byKarlsson et al. found that there is no general trend for nanosized particles to be more toxic than micrometer-sized, but rather it depends on the mineral type. Based on this information we can assume that the magnetic fraction in the urban dust can also play role in generation of negative effects on human health.
 Assuming initially that toxicological effects of iron oxides are simply related to concentration changes, we examine the existence of relationships between mass-specific magnetic susceptibility (χ) of both indoor and outdoor dusts, and mortality rates caused by respiratory and cardiovascular diseases (Figure 12). The absence of such dependence is most probably related to different contribution of various sources of magnetic particles in the sampled dust. To take it into account, we examined the possibility of “normalization” of mass specific susceptibility to a parameter, which would represent the regional urban background. Magnetic susceptibility of natural soils, spread in the surroundings of the corresponding city cannot be used as a “background” signal for several reasons: i) in some places wide variety of soil types with contrasting magnetic susceptibilities are present. For example in Sofia valley, soil types range from weakly magnetic Vertisols up to strongly magnetic Cambisols; ii) there are in general very different proportions of soil-derived material with respect to the anthropogenic fraction in different dust samples. Therefore, normalizing to a value, corresponding to 100% topsoil, will not help to discriminate the two contributions. This leads us to use another approach. We calculate the ratio of indoor to outdoor susceptibilities (χindoor/χoutdoor). This ratio can be considered as analogous to the indoor/outdoor (I/O) ratios frequently reported for the PM concentrations and various chemical elements. Different values of the I/O parameter are reported for various grain sizes and different chemical elements [Jones et al., 2000; Chen and Zhao, 2011]. In the absence of indoor sources, I/O ratio is less or equal to 1. Concerning the origin of magnetic susceptibility signal of settled dust, similar considerations can be applied. The magnetic susceptibility of outdoor dust reflects the mixed contribution of: i) soil-derived strongly magnetic dust particles and ii) anthropogenic (industrial, traffic-related, fossil-fuel etc.) particles. Indoor dust in the present study has mainly an ambient origin, i.e., it contains magnetic particles of similar origin, but with grain sizes selectively oriented toward finer sizes. This fining of infiltrated and penetrated magnetic fraction is due to the fact that coarse particles penetrate with more difficulty indoors due to gravitational differentiation. This implies that most of the coarse dust particles will be settled outside, while only finer grained particles could penetrate indoors. Close proximity of the sampled locations, which correspond to an outdoor and indoor sampling point was used to assure that there will be no significant differences in the main dust source. In case of equal conditions for penetration, infiltration and re-suspension of dust particles, the amount of ambient (outdoor) particles indoor will depend on the size distribution and amount of anthropogenically derived magnetic particles. Thus, high values of the ratio (χindoor/χoutdoor) will be obtained if larger quantities of outdoor particles penetrate indoors, and vice versa. Combustion (both industrial and traffic-related) is the primary source of most carcinogenic and toxicogenic particulates in ambient air, and it generates particles in the fine grained range (PM10 in general). Consequently, the amount of combustion-originating magnetic particles which penetrate indoors will be mostly responsible forχindoor. In favor of differential infiltration behavior of dust particles with different origin are the results presented by Meng et al. , who showed that infiltration factors for PM2.5 originating from primary combustion, secondary formation and mechanical formation differ significantly, being much higher for the first two groups of particles. The obtained range of variations of the ratio χindoor/χoutdoor may be compared to the widely studied PM10 mass indoor/outdoor (I/O) ratio [Thatcher and Layton, 1995; Jones et al., 2000; Liao et al., 2003, 2004]. This similarity suggests that magnetic fraction in indoor and outdoor dusts obeys the same behavior as the major mineral dust components, contributing to the amount of PM.
 Comparison between mean annual values of χindoor/χoutdoor per city with the corresponding data on mortality caused by respiratory and cardiovascular diseases (Figure 13) reveals the existence of a linear correlation between the ratio χindoor/χoutdoor and the number of people dying as a result of cardiovascular diseases. The obtained correlation coefficient R2 of 0.82 at p < 0.05 significance level refers to the classical calculation without accounting for the error bars on χ-ratio. A t-test is applied in order to test the significance of the regression, i.e., hypothesis that the slope of the regression equation (b2) is equal to zero. Rejection of the null hypothesis has been obtained, suggesting that the fitted regression model is of value in explaining variations in the observations. Confidence Intervals on Regression Coefficients (intercept b1 and slope b2) has been determined as well [Sen and Srivastava, 1990; Montgomery et al., 2001]. Figure 14a shows the estimated confidence area determined by b1 and b2, which is significantly different from the line of zero slope, thus proving the significance of the regression. Equations used for statistical tests on the significance of the linear regression are given in the auxiliary material. Taking into account the experimental errors in χ-ratio, plot of the highest posterior density within the 95% envelope (Figure 14b) shows that there is still significant relationship between magnetic ratio and cardiovascular mortality.
 The obtained correlation could be tentatively explained by the following considerations. The ratio χindoor/χoutdoorprimarily depends on the amount of outdoor magnetic particles, which were able to penetrate indoors. These are, as discussed above, enriched in finer material of anthropogenic origin. The greater number of particles of this origin penetrating indoors, the higher will be the risk for PM-related health problems. Thus, higher values ofχindoor/χoutdoor signify an increased anthropogenic pollution level indoors. Due to the long time spent by people indoors [Klepeis et al., 2001], proportional increases in mortality caused by cardiovascular diseases with increased values of χindoor/χoutdoor should be expected.. Now an important issue is why and how increases in atmospheric pollution are more relevant to mortality caused by cardiovascular diseases (myocardial infarction, heart failure, and fatal arrhythmias) and not so significant in respiratory diseases. The traditional view that the lungs are most affected by air pollution is not supported by the epidemiological evidence during the last 15 years [Johnson, 2004]. Pope et al. [2004, p. 71] also report that “fine particulate air pollution is a risk factor for cause-specific cardiovascular disease mortality via mechanisms that likely include pulmonary and systemic inflammation, accelerated atherosclerosis, and altered cardiac autonomic function. Although smoking is a much larger risk factor for cardiovascular disease mortality, exposure to fine PM imposes additional effects that seem to be at least additive to if not synergistic with smoking.” A recent review [de Kok et al., 2006] shows that toxicity of ambient and traffic-related PM is not always proportional to the intensity of the traffic but rather is source-dependent. This implies that our approach in using close pairs of indoor – outdoor sites for estimation of health impact of urban dust is highly relevant.
 The obtained scatter in the plot of the ratio χindoor/χoutdoor versus mortality caused by respiratory diseases (Figure 13a) could be due to the fact that respiratory tract is much more influenced by tobacco smoke, the latter being among the primary causes of lethal effects for vulnerable population [Hecht, 1999]. Data points which deviate the most from tentative linear dependence are data for Burgas (2009 and 2010), Russe and Plovdiv – 2010 data.