- Email: [email protected]
Magnetic signature of daily sampled urban atmospheric particles
ARTICLE IN PRESS
Atmospheric Environment 37 (2003) 4163–4169
Magnetic signature of daily sampled urban atmospheric particles Adrian R. Muxworthya,*, Jurgen . Matzkab, Alfonso Ferna! ndez Davilab, Nikolai Petersenb a
School of GeoSciences, Grant Institute, School of GeoSciences, University of Edinburgh, West Mains Road, Edinburgh EH9 3JW, UK b Geophysics, Department of Earth and Environmental Sciences, Ludwig-Maximilians Universitat, . Theresienstrasse 41, Munich 80333, Germany Received 28 April 2003; received in revised form 22 May 2003; accepted 28 May 2003
Abstract The magnetic signature of two sets of daily sampled particulate matter (PM) collected in Munich, Germany, were examined and compared to variations in other pollution data and meteorological data using principal component analysis. The magnetic signature arising from the magnetic minerals in the PM was examined using a fast and highly sensitive magnetic remanence measurement. The longest data set studied was 160 days, significantly longer than that of similar magnetic PM studies improving the statistical robustness. It was found that the variations in the massdependent magnetic parameters displayed a complicated relationship governed by both the meteorological conditions and the PM loading rate, whereas mineralogy/grain-size-dependent magnetic parameters displayed little variation. A six-fold increase in the number of vehicles passing the sampling locations only doubled the magnetic remanence of the samples, suggesting that the measured magnetic signature is in addition strongly influenced by dispersion rates. At both localities the saturation isothermal remanent magnetisation (SIRM) was found to be strongly correlated with the PM mass, and it is suggested that measuring SIRM as a proxy for PM monitoring is a viable alternative to magnetic susceptibility when the samples are magnetically too weak. The signal was found to be dominated by magnetite-like grains less than 100 nm in diameter which is thought to be derived primarily from vehicles. Such small grains are known to be particularly dangerous to humans. There was also evidence to suggest from magnetic stability parameters that the magnetite-like grains were covered with an oxidised rim. The concentration of magnetic PM was in the range of 0.3– 0.5% by mass. r 2003 Elsevier Ltd. All rights reserved. Keywords: Magnetic measurements; Iron oxide; Vehicle pollution; Urban PM
1. Introduction Urban atmospheric particles are known to adversely effect human health (Harrison and Yin, 2000; Zhu et al., 2001). Hence, there is a need to identify both the constituents and atmospheric transport pathways of urban particulate matter (PM). Due to the high content of magnetic minerals in urban PM, the application of *Corresponding author. Fax: +44-131-668-3184. E-mail address: [email protected] (A.R. Muxworthy).
magnetic techniques to rapidly assess urban PM is increasingly being seen as a viable approach (e.g., Matzka and Maher, 1999; Shu et al., 2000, 2001). Magnetic minerals are common amongst pollution PM, with bulk iron content found to constitute 1% of urban atmospheric PM (Department of the Environment, 1996), and iron oxides and hydroxides typically contributing to 10–70% of the bulk iron content (Dedik et al., 1992; Weber et al., 2000). Iron impurities in fossil fuels convert on combustion to magnetic iron oxides, i.e., magnetite, hematite or a mixture of both. Due to its combustion origin, magnetic PM is not only dangerous
1352-2310/03/$ - see front matter r 2003 Elsevier Ltd. All rights reserved. doi:10.1016/S1352-2310(03)00500-4
ARTICLE IN PRESS 4164
A.R. Muxworthy et al. / Atmospheric Environment 37 (2003) 4163–4169
in itself (Phumala et al., 1999; Garcon et al., 2000, 2001), but is also associated with other hazardous pollutants which are injected into the atmosphere during combustion (Morris et al., 1995; Muxworthy et al., 2001). PM in vehicle exhaust is dominated by particles o 10 mm, i.e., PM10 (Kim et al., 2001). In addition to direct combustion derived particles in urban environments vehicles produce other magnetic PM via abrasion/corrosion in . particular brake wear (Olson and Skogerboe, 1975; Osterle et al., 2001), which also contains particles o 10 mm (McCrone and Delly, 1973; Gillies and Gertler, 2000). The relationship between the magnetic signature of PM and other pollution data, e.g., PM mass and gaseous pollution products, is still not fully understood. There have been a limited number of studies which have tried to resolve this relationship by examining time series (Morris et al., 1995; Muxworthy et al., 2001). These studies have found that the magnetic signature displays a complicated behaviour, which is related to other pollution data, but is also strongly effected by meteorological variations, and in addition is highly site dependent. Morris et al. (1995) found a strong correlation between the magnetic susceptibility of daily sampled PM and pollutants such as NO2. However, susceptibility yields only limited information about the magnetic content if no other magnetic information is known. Muxworthy et al. (2001) extended the work of Morris et al. (1995) by measuring magnetic hysteresis of urban PM. Magnetic hysteresis data is more descriptive than susceptibility data as it provides more information about mineralogy, grain size and concentration (Dunlop . and Ozdemir, 1997), which in turn provides more information about the source and characteristics of the PM. Muxworthy et al. (2001) found strong correlations between grain-size-dependent magnetic parameters and the relative humidity. However in their study, due to the insensitivity of the magnetometer used to measure the magnetic hysteresis, PM samples were measured in weekly batches. This averaging removed much of the signal variation and decreased the quality of the data. In addition, due to the rather long measuring time of hysteresis it was not possible to measure long time series. The initial proposal for this study was to advance and improve on the studies of Morris et al. (1995) and Muxworthy et al. (2001), by measuring time series of PM using magnetic techniques which allowed for: daily samples to be measured, changes in magnetic stability (controlled by grain-size and mineralogy) and total magnetic content to be determined, and for long time series to be measured to reduce statistical error. In Morris et al. (1995) and in particularly in Muxworthy et al. (2001) the number of data points used in the statistical analysis was rather low; as a rough guideline the number of data points should be at least B100. To meet these requirements magnetic remanence mea-
surements were made. Although remanences have significantly smaller magnetic signals than in-field measurements, e.g., magnetic hysteresis or magnetic susceptibility, it is possible to measure remanences by using superconducting quantum interference device magnetometers which are highly sensitive and allow for smaller samples to be measured. A set of experiments were designed which combined remanence inducement and partial demagnetisation to assess magnetic stability, and had a relatively short measurement time.
2. Samples and methodology PM samples were collected by the Bayerisches Landesamt fur . Umweltschutz (BLFU, Bavarian Department for the Environment) at the permanent air monitoring stations at Luise-Kiesselbach-Platz (LKP, EU code DEBY085) and WestendstraXe (WS, EU code DEBY045). PM10 was collected using Eberline FH 62I-N dust sampler over 24-h periods. At WS the filter samples were measured between 1 April 2000–7 September 2000 (160 days) and at LKP 24 August 2000–31 October 2000 (69 days). PM10 concentration was measured by its absorption of beta rays and the sampled air volume per time unit was kept constant, taking into account air temperature and pressure. The WS station is 4 km west of the city centre and classified as being in the city (Fig. 1), while LKP is 3 km south of WS and is classified as being in the city suburbs (BLFU, 1998). The prevailing wind direction for Munich is westerly. The daily average number of cars for the period 1990–1995 was 20,000 per day at WS and
48˚16′38
48˚14′08
N DWD
48˚11′38
WS 48˚08′58
LKW
CENTRE
48˚06′28
48˚03′58
PREVAILING WIND DIRECTION
5 km
48˚01′28 11˚25′01 11˚27′31 11˚30′01 11˚32′31 11˚35′01 11˚37′31 11˚40′31
Fig. 1. Map of Munich showing the locations of the two pollution monitoring sites LKW (DEBY085) and WS (DEBY045), and the Deutscher Wetter Dienst (DWD, German Weather Service).
ARTICLE IN PRESS A.R. Muxworthy et al. / Atmospheric Environment 37 (2003) 4163–4169
Table 1 Summary of the daily mean MASS, SIRM and SIRM/MASS for the two data sets in this examined study WS
MASS (mg) SIRM ( 10 9 A m2) SIRM/MASS (mA m2/kg)
LKP
Mean
s
Mean
s
0.6 18.9 32.1
0.2 0.2 0.1
0.8 36.5 44.3
0.2 0.4 0.1
The daily means were calculated using a log-normal distributions, and converted back to a linear scale. s is the standard deviation of the normal distribution, and is not symmetric about the mean on a linear scale. At WS the number of measurements used to calculate s was 160, and at LKP 69.
SIRM
normalised remanence
117,000 per day at LKP (BLFU, 1998). The WS sampling station is situated 6 m from a road, whilst the LKP station is located 2 m from the road. At WS there is a tram line running along the road. Both samplers are placed 4 m above the ground. In addition to being located close to a road, WS is located approximately 50 m from a workshop for public transport buses and trams, which could potentially contribute magnetic PM not necessarily associated with gaseous pollution products. At both sites in addition to PM10 concentration, CO, NO, NOx (NO+NO2) and SO2 were measured. The sampling techniques and instrumentation for this data is fully described in BLFU (1998). Meteorological data was provided by the Deutscher Wetter Dienst (DWD, German Weather Service). Air temperature, relative humidity, wind speed and rainfall were measured at the meteorological station ‘‘Munchen . Stadt’’ of the DWD, which is located B4 km NNE from WS (Fig. 1). Wind direction data was not considered, because of the effects of localised wind channelling. This channelling effect may also effect the quality of the wind speed data. The dust filters provided by the BLFU were 4 cm broad bands containing series of circular dust spots with 2 cm diameter. Each spot contains the dust sampled from 24 m3 of air during a 24 h sampling interval starting at 0:00 local time. Razor blades were used to cut out the dust covered filter material. These circular filter pieces were folded several times to contain the dust inside the filter material and firmly fixed in a 1 cm long piece of plastic drinking straw. This sample preparation allowed manipulation without loosing any of the dust and provided the necessary mechanical stability for magnetic dust particles to carry a stable remanence direction. The samples were given a saturation isothermal remanent magnetisation (SIRM) in a field of 1 T with an electromagnet. From hysteresis measurements on similar PM10 samples collected in Munich, such a field is known to be strong enough to induce a SIRM (Matzka, 1997; Muxworthy et al., 2001). The samples were demagnetised using a 2 G alternating field (AF) demagnetiser. After each treatment, the straw was measured in two positions with a cryogenic magnetometer in a fieldfree room. To test the reliability of this method, the remanence carried by empty drinking straws and blank filter paper were measured. The SIRM of the sample holder plus the empty straw gives a magnetic moment of B6 10 11 A m2 close to the detection limit of the cryogenic magnetometer. The blank filter paper gives values close to 1.4 10 9 A m2, some 13 times smaller than the signal from average SIRM signal of WS (19 10 9 A m2) and 25 times smaller than the signal from LKP (36 10 9 A m2) (Table 1). In general the signal of the filter paper was small compared to that of the dust, and was considered negligible in this study.
4165
1.0
AF10
0.8
37 nm
AF25
0.6 76 nm
0.4
215 nm
0.2 0.0 0
10
20
30 40 AF field (mT)
50
60
Fig. 2. AF demagnetisation curve for a test sample from WS (24 August 2000). Also shown are the AF demagnetisation curves for three synthetic magnetite samples with mean grain sizes of 37, 76 and 215 nm, from Dunlop and Argyle (1991).
To measure a sufficient number of samples for statistical analysis, a fast measuring scheme was adopted. Samples were produced from the filter bands, magnetically treated and measured in batches of eight. We reduced the information content of each AF demagnetisation curve to two representative parameters; firstly a SIRM, which has been found to be representative of the total magnetic content of the sample. Secondly the ratio AF25/AF10 which is the ratio of SIRM AF demagnetised with a 25 mT field divided by SIRM AF demagnetised with a 10 mT field (see Fig. 2). This ratio assesses the magnetic stability of the sample against AF demagnetisation. If the magnetic signal is dominated by one mineral, then AF25/AF10 is controlled only by magnetic domain state which is directly related . to grain size or internal structure (Dunlop and Ozdemir, 1997). We considered this ratio instead of AF25/SIRM, because we wanted to assess variations in only higher coercive fractions of remanence. Thermomagnetic curves measured by Muxworthy et al. (2001) found that these samples had only one significant phase which was magnetite-like.
ARTICLE IN PRESS A.R. Muxworthy et al. / Atmospheric Environment 37 (2003) 4163–4169
4166
that the PM10 has, like NO, a short residency time in the air.
3. Results 3.1. Magnetic measurements
3.2. Daily measurements The normalised day-of-week averages for the magnetic and pollution data for WS and LKP are plotted in Fig. 3. All the data sets at WS and LKP, with the exception of AF25/AF10, display low-weekend values (Sunday lower than Saturday). The Thursday low in LKP data set is attributed to the very high average rainfall on Thursdays, which occurred during the relatively short time series (69 days). The PM data, i.e., PM10 and SIRM, appear to be particularly affected by the precipitation. AF25/AF10 displays virtually no week–day variation, suggesting only one magnetic source. That the PM10 and SIRM data display very similar daily behaviour to the NO data, suggests that the magnetic signature has both the same source as NO and
3.3. Principal component factor analysis The magnetic results were compared to the pollution and meteorological data using principal component factor analysis (PCFA). The magnetic variables and PM10 were transformed using logarithmic base 10 function in a similar manner to that described elsewhere in the literature (e.g., Morris et al., 1995; Urbat et al., 1999). This transformation accommodates the lognormal distribution of these parameters (Kim et al., 2001) . The meteorological data (precipitation, temperature, wind speed and relative humidity) were not measured at the same locations at the pollution data; however, they were taken as global, that is, there would be little spatial variation within Munich in the variance when averaged daily. The variable most likely not to be global being the precipitation. To simplify and reduce the matrix during PCFA, only representative meteorological, pollution and magnetic data was selected. The meteorological data was represented by precipitation, temperature, wind speed and relative humidity, the pollution data by CO, NO and PM10. Both NOx and SO2 were strongly correlated with both CO and NO. The magnetic data was represented by SIRM and AF25/AF10. Bi-plots of some combinations of these parameters are shown in Fig. 4.
1.5
1.0
1.0
0.5
0.5
NO
SIRM
1.5
SIRM (×10-7Am2)
1
+ ++ + + + + + ++ ++ + +++++++++++++ +++ ++++++++++ ++ +++ ++ + + + + + + + ++ ++++ ++++++++++ + + +++ + +++++ +++ + ++++ +++ +++++++++++ ++ +++ ++ ++ + ++++ ++ ++++ + + + ++ + + + + ++
0.1 +
0.1
1
(a) 0.0
+ + ++ + ++ + + + + ++ ++ + + + + + +++ ++ + ++ +++++++++ +++++++ ++++ ++++ + + ++++ + + + + + ++ ++++++++ + + ++ + + + ++++ ++++++++++ + ++ + ++ + + ++ + + ++ ++ ++++++++ + +++ +++ + + + + ++++ + + + ++ + + +++ + +
0.1
+
0.0
(b)
MASS (mg)
0.5
1.0
1.5
CO (mgm-3)
0.0
(c)
0.8 1
1.5
PM 10
1.0
0.5
0.5
0.0
0.0
SIRM (×10-7Am2)
1.0
AF25 /AF10
1.5
AF25 /AF10
(a)
1
+ WS LKP
SIRM (×10-7Am2)
Complete AF demagnetisation curves were measured for a representative selection of samples. In Fig. 2 a typical AF demagnetisation curve is shown together with standard curves for synthetic, sized magnetite samples. Several features can be seen, firstly at lowfields the curve for the PM10 sample lies near that of the 76 nm magnetite sample. As the field increases the PM10 sample’s curve decreases less rapidly than those of the synthetic magnetite samples, and is magnetically more stable.
0.7 +
0.6
+ + ++++++ + + +++++ ++++ + +++++ +++++ ++++ ++ ++++ ++++ ++++++++++++ ++++++++++++++ ++ + ++ + + +++++++ + +++ + ++++ +++ +++ +++++ ++++ + ++ + ++ ++++ ++++ + + + + +
+ +++ ++ + + ++++ ++ + +++ + ++ + ++ +++++++++ + ++ ++++ ++ ++++++++++++ +++ +++ ++ + + +++ + +++++++ ++ ++ + + ++ + + + ++ ++++++++++++++++ ++ + ++++ ++ +++ +++++ +++ ++ +++ + ++ ++++
0.1
+
0.5
(b)
Sa Su M T W Th F week day
(d)
Sa Su M T W Th F week day
Fig. 3. Normalised day-of-week averages from LKP (light grey) and WS (dark grey) for (a) CO, (b) PM10, (c) SIRM and (d) the ratio AF25/AF10. The parameters are normalised by the total average value.
(c)
0.1 SIRM (×10-7Am2)
1
0
(d)
2 4 wind speed (ms-1)
Fig. 4. Bi-plots of various parameters plotted against each for data from WS and LKP: (a) SIRM versus PM10; (b) SIRM against CO; (c) AF25/AF10 against SIRM; and (d) SIRM versus wind speed. SIRM versus NO (not shown) displays similar trends to (b).
ARTICLE IN PRESS A.R. Muxworthy et al. / Atmospheric Environment 37 (2003) 4163–4169 1.0 CO
WS
NO
0.5
factor 2
RAIN
PM 10
RH 0.0 AF25 /AF10 SPEED
TEMP
SIRM
-0.5
-1.0
(a) 1.0 CO
LKP
NO
PM10
SIRM
0.5
factor 2
RH
AF25 /AF10
0.0 TEMP RAIN -0.5 SPEED
-1.0 -1.0
(b)
-0.5
0.0
0.5
1.0
factor 1
Fig. 5. Factor loading plots derived from PCFA for the magnetic data (SIRM and AF25/AF10) sequences from (a) WS and (b) LKP, with representative pollution and meteorological data. MASS is the log10 of the mass of the samples, RAIN is the precipitation, SPEED is the wind speed, RH is the relative humidity and the pollutants CO and NO.
PM10 and SIRM display a strong linear dependency with LKP having higher mean values than WS (Table 1 and Fig. 4a). Extrapolating the linear regression on to the axes, the trend passed within error through the origin for both LKP and WS, implying that there was no non-magnetic source contributing to the PM10 signal at either locality. The relationship between SIRM and the gaseous pollution products, e.g., CO etc., is less transparent (Fig. 4b); the data from WS is scattered, whereas LKP displays a stronger linear relationship. There is little variation in AF25/AF10 as SIRM increases (Fig. 4c), suggesting a consistent loading material. The relationship between SIRM and the meteorological data is less clear (Figs. 4d). Increases in wind speed reduce SIRM, suggesting that the source of SIRM is local. This effect is more pronounced in the LKP data. The factor plot of the pollution, meteorological and magnetic data for WS is shown in component space for the first two components in Fig. 5a. The first two factors contribute only 56% of the variance, while the third
4167
component contributes a further 13% and the fourth 11%, however, the main relationships are still visible. PM10 and SIRM plot close to each other as they are strongly correlated with a Spearman’s r significant at the 0.01 level. These two variables are correlated with temperature which is anti-correlated with the precipitation, relative humidity and wind speed. The pollution data represented by CO and NO display a strong correlation with each other, but display no strong correlation with any other variable. The lack of a strong relationship is also seen in Fig. 4b. AF25/AF10 is displays little variation, and plots between the origin and the temperature. In Fig. 5b, the pollution, meteorological and magnetic data for LKP are plotted in component space for the first two factors representing 60% of the variance. The third factor contributed only 14% of the variance. On the factor 1 axis the meteorological data, with the exception of precipitation, plot in similar position as in Fig. 5a, and CO and NO plot on the factor 2 axis. The wind speed data is shifted slightly and lies closer to the factor 2 axis. The sampling period for the LKP data is different to that of WS. PM10 and SIRM are still closely correlated with each other (ro0:01), however, both parameters now lie closer to CO and NO. There is a stronger correlation with the gaseous pollution data for SIRM (ro0:01) than at WS. The wind speed data is strongly anti-correlated with PM10 and SIRM (ro0:01).
4. Discussion For low-fields the AF demagnetisation curves suggest that the magnetite-like grains are o100 nm, however, as the AF increases the PM10 samples display a higher magnetic stability than for the synthetic samples (Fig. 2). The high stability may indicate the presence of a second phase. This phase is most likely be to an oxidised surface layer; a maghemite rim or near-maghemite rim on the magnetite particles. Due to differences in lattice spacing between magnetite and maghemite, such rims induce a surface stress which effectively increase the stability of the particles giving rise to underestimates for the mean magnetic grain size. Such oxidation processes occur initially very rapidly, however, once the rim is formed further oxidation of the magnetite core occurs very slowly. This may also explain the difference between the mean grain size found in this study compared to previous studies for PM10 from Munich which found a magnetite-like phase with a mean grain size 200–500 nm (Muxworthy et al., 2001). There are clear differences between the two locations studied. The correlation between the magnetic parameters, e.g., SIRM, and the gaseous pollution products, e.g., CO and NO, is site dependent. At LKP SIRM and PM10 were closely related to CO and NO (Figs. 4b and 5b), which can
ARTICLE IN PRESS 4168
A.R. Muxworthy et al. / Atmospheric Environment 37 (2003) 4163–4169
also be seen in the averaged daily trends (Fig. 3). At WS the correlation between these parameters was lower. Considering that the nearly six times as many cars pass LKP than WS, it is seen that the effect of increased traffic loading produces a closer correlation between pollution and magnetic data, i.e., the relative influence of the meteorological effects decreases as the pollution loading increases. However, the exact relationship is complicated and non-linear; attempts to produce a general non-sitespecific linear regression model were unsuccessful. Are the differences between the mean values for SIRM and PM10 from the two localities (Table 1) due to spatial or temporal reasons? To assess this, we consider the ratios of the mean values for the data from the 15 days where the two time series overlap. The ratios for LKP over WS for SIRM and PM10, are respectively B2 and B1.4, implying that the differences are due to spatial variation. It is possible to determine the magnetic content as a percentage by mass if we assume two things; firstly that magnetite is the primary magnetic mineral and secondly that the saturation magnetisation (MS ) can be determined by taking the mean SIRM=MS ratio B0.11 from Muxworthy et al. (2001). By extrapolating from SIRM to determine MS ; we are also incorporating the nonremanence carrying grains, i.e., superparamagnetic grains. For WS the percentage by mass of magnetite is determined to be 0.32% using the data in Table 1, which is identical to that reported for PM70 in Muxworthy et al. (2001), and 0.44% for LKP. Again it is seen that a sixfold increase in vehicles produces only a relatively small increase in the magnetic content of the sample, implying that either vehicles are not large contributors to the magnetic signal, or more likely magnetic PM10 distributes quickly to give an average background level. Matzka and Maher (1999), Hoffmann et al. (1999) and Moreno et al. (2003) found that roadside pollution was mostly deposited within the first few metres from the road suggesting that the pollution is not evenly distributed. Muxworthy et al. (2002) found no significant difference between PM collected at 15 and B150 m from a road in central Munich. Of interest to palaeoclimate studies is how the magnetic signatures are effected by meteorological conditions. Increases in wind speed, precipitation and relative humidity all decrease PM10 and SIRM, though for different reasons. Increases in wind speed are likely to move PM to other areas away from the source station. Relative humidity and precipitation act differently by reducing residency time (Muxworthy et al., 2001). That is, precipitation and relative humidity increase local deposition rates, whereas increasing wind speed decreases it. The reduction of PM10 and SIRM with increasing relative humidity is attributed to hygroscope interaction between SO2 absorbed on the surface of magnetite
grains with water vapour within the air as discussed by Muxworthy et al. (2001). However, Muxworthy et al. (2001) noticed a subtler interaction between the magnetic grain size distribution and the relative humidity; from hysteresis measurements, they found that the smallest grains were preferentially removed by increases in relative humidity. No such relationship was found between AF25/AF10 and the relative humidity in this study. It is suggested that the AF25/AF10 is less sensitive than hysteresis measurements at determining changes in grain size and mineralogy. The relationship between SIRM and wind speed (Fig. 4d) is non-linear. There appears to be a cut-off wind speed above which it is not possible to obtain higher SIRM values at these two locations. It is uncertain if this is a real effect or a technical sampling error. If a real effect, it is likely that the critical wind speed is dependent on the loading rate and location.
5. Conclusions Detailed analysis of the behaviour of the magnetic signature of daily sampled suspended particulate matter (PM) collected in Munich in 2000 reveals that the magnetic signature displays a complicated non-linear relationship with both gaseous pollutants and the meteorological data. The saturation magnetic remanence was found to be strongly correlated with PM10 suggesting that the behaviour of PM10 is adequately described by the behaviour of SIRM. This is of course highly site dependent and assumes that PM loading is from a single homogenous magnetic source. Muxworthy et al. (2001) found that at another location in Munich (Pasing) that this relationship did not hold. However, at these two localities it is thought that there is only one consistent magnetic source, i.e., no non-magnetic sources, and that this is most likely to be derived from vehicles. The magnetic signature appears to be dominated by a magnetite-like grains o100 nm in diameter, probably with an oxidised rim. Such small grains are known to be particularly dangerous to humans as they can be inhaled deep into the lungs (Smith and Aust, 1997; Donaldson et al., 1998). At the two locations there was found to be little variation in the mineralogy and/or grain size of the magnetic PM. Measuring remanence rather than the more standard magnetic susceptibility, does not greatly increase measurement time, however, because of increased sensitivity it allows for significantly smaller samples to be measured. Moreno et al. (2003) found a strong correlation between magnetic susceptibility and SIRM in their study of PM trapped on leaves. It is suggested therefore that in cases where the samples are too weak to measure the susceptibility, SIRM studies provide a viable alternative.
ARTICLE IN PRESS A.R. Muxworthy et al. / Atmospheric Environment 37 (2003) 4163–4169
Acknowledgements The authors thank Mr. Weber of the Bayerisches Landesamt fur . Umweltschutz for providing both the filter samples and the pollution data. This work was funded by the European Union (Contract no. ERBFMXCT 98-0247), as part of the European Network for Mineral Magnetic Studies of Environmental Problems, and NERC grant NER/A/S/2001/00539.
References Bayerisches Landesamt fur . Umweltschutz (BLFU), 1998. Lufthygienischer Jahresbeircht 1997. Bayerisches Landesamt fur . Umweltschutz, Munchen. . Dedik, A.N., Hoffmann, P., Ensling, J., 1992. Chemical characterization of iron in atmospheric aerosols. Atmospheric Environment 26A (14), 2545–2548. Department of the Environment, 1996. Airborne particulate matter in the United Kingdom, Third Report of the Air Quality of Urban Air Review Group, Department of the Environment, UK. Donaldson, K., Li, X.Y., MacNee, W., 1998. Ultrafine (nanometre) particle mediated lung injury. Journal of Aerosol Science 29, 553–560. Dunlop, D.J., Argyle, K.S., 1991. Separating multidomain and remanence induced in pseudo-single-domain magnetites (215–540 nm) by low-temperature demagnetization. Journal of Geophysical Research 96 (2), 2007–2017. . . 1997. Rock Magnetism: FundaDunlop, D.J., Ozdemir, O., mentals and Frontiers. Cambridge Studies in Magnetism. Cambridge University Press, Cambridge, 573pp. Garcon, G., Shirali, P., Garry, S., Fontaine, M., Zerimech, F., Martin, A., Hannothiaux, M.H., 2000. Polycyclic aromatic hydrocarbon coated onto Fe2O3 particles: assessment of cellular membrane damage and antioxidant system disruption in human epithelial lung cells (L132) in culture. Toxicology Letters 117 (1–2), 25–35. Garcon, G., Gosset, P., Garry, S., Marez, T., Hannothiaux, M.H., Shirali, P., 2001. Pulmonary induction of proinflammatory mediators following the rat exposure to benzo(a)pyrene-coated onto Fe2O3 particles. Toxicology Letters 121 (2), 107–117. Gillies, J.A., Gertler, A.W., 2000. Comparison and evaluation of chemically speciated mobile source PM2.5 particulate matter profiles. Journal of the Air and Waste Management Association 50 (8), 1459–1480. Harrison, R.M., Yin, J.X., 2000. Particulate matter in the atmosphere: which particle properties are important for its effects on health? Science of the Total Environment 249 (1–3), 85–101. Hoffmann, V., Knab, M., Appel, E., 1999. Magnetic susceptibility mapping of roadside pollution. Journal of Geochemical Exploration 66, 313–326. Kim, W.-K., Kim, S.H., Lee, D.W., Lee, S., Lim, C.S., Ryu, J.H., 2001. Size analysis of automobile soot particles using field-flow fractionation. Environmental Science and Technology 35, 1005–1012.
4169
Matzka, J., 1997. Magetische, elektronenmikroskopische und lichtmikroskopische Untersuchungen an St.auben und Aschen sowie an einzelnen Aschepartikeln. Universit.at M.unchen. Matzka, J., Maher, B.A., 1999. Magnetic biomonitoring of roadside tree leaves: identification of spatial and temporal variations in vehicle- derived particulates. Atmospheric Environment 33, 4565–4569. McCrone, W.C., Delly, J.G., 1973. The Particle Atlas, 2nd Edition. Ann Arbor Science Publishers, Ann Arbor. Moreno, E., Sagnotti, L., Dinars-Turell, J., Winkler, A., Cascella, A., 2003. Biomonitoring of traffic air pollution in rome using magnetic properties of tree leaves. Atmospheric Environment 37 (21), 2967–2977. Morris, W.A., Versteeg, J.K., Bryant, D.W., Legzdins, A.E., McCarry, B.E., Marvin, C.H., 1995. Preliminary comparisons between mutagenicity and magnetic-susceptibility of respirable airborne particulate. Atmospheric Environment 29, 3441–3450. Muxworthy, A.R., Matzka, J., Petersen, N., 2001. Comparison of magnetic parameters of urban atmospheric particulate matter with pollution and meteorological data. Atmospheric Environment 35 (26), 4379–4386. Muxworthy, A.R., Schmidbauer, E., Petersen, N., 2002. Magnetic properties and M.ossbauer spectra of urban atmospheric particulate matter: a case study from Munich, Germany. Geophysical Journal International 150 (2), 558–570. Olson, K.W., Skogerboe, R.K., 1975. Identification of soil lead compounds from automotive sources. Environmental Science and Technology 9, 227–230. . Osterle, W., Griepentrog, M., Gross, Th., Urban, I., 2001. Chemical and microstructural changes induced by friction and wear of brakes. Wear 251, 1469–1476. Phumala, N., Ide, T., Utsumi, H., 1999. Non-invasive evaluation of in vivo free radical reactions catalyzed by iron using in vivo ESR spectroscopy. Free Radical Biology and Medicine 26 (9–10), 1209–1217. Shu, J., Dearing, J.A., Morse, A.P., Yu, L.Z., Li, C.Y., 2000. Magnetic properties of daily sampled total suspended particulates in Shanghai. Environmental Science and Technology 34 (12), 2393–2400. Shu, J., Dearing, J.A., Morse, A.P., Yu, L.Z., Yuan, N., 2001. Determining the sources of atmospheric particles in Shanghai, China, from magnetic and geochemical properties. Atmospheric Environment 35 (15), 2615–2625. Smith, K.R., Aust, A.E., 1997. Mobilization of iron from urban particulates leads to generation of reactive oxygen species in vitro and induction of ferritin synthesis in human lung epithelial cells. Chemical Research in Toxicology 10, 828–834. Urbat, M., Dekkers, M.J., Vriend, S.P., 1999. The isolation of diagenic groups in marine sediments using fuzzy c-means cluster analysis. In: Tarling, D.H., Turner, P. (Eds.), Palaeomagnetism and Diagenesis in Sediments, Vol. 151. Geological Society, London, pp. 85–93. Weber, S., Hoffmann, P., Ensling, J., Dedik, A.N., Weinbruch, S., Miehe, G., Gutlich, P., Ortner, H.M., 2000. Characterization of iron compounds from urban and rural aerosol sources. Journal of Aerosol Science 31 (8), 987–997. Zhu, R.X., Shi, C.D., Suchy, V., Zeman, A., Guo, B., Pan, Y.X., 2001. Magnetic properties and paleoclimatic implications of loess–paleosol sequences of Czech Republic. Science in China Series D-Earth Sciences 44 (5), 385–394.
Atmospheric Environment 37 (2003) 4163–4169
Magnetic signature of daily sampled urban atmospheric particles Adrian R. Muxworthya,*, Jurgen . Matzkab, Alfonso Ferna! ndez Davilab, Nikolai Petersenb a
School of GeoSciences, Grant Institute, School of GeoSciences, University of Edinburgh, West Mains Road, Edinburgh EH9 3JW, UK b Geophysics, Department of Earth and Environmental Sciences, Ludwig-Maximilians Universitat, . Theresienstrasse 41, Munich 80333, Germany Received 28 April 2003; received in revised form 22 May 2003; accepted 28 May 2003
Abstract The magnetic signature of two sets of daily sampled particulate matter (PM) collected in Munich, Germany, were examined and compared to variations in other pollution data and meteorological data using principal component analysis. The magnetic signature arising from the magnetic minerals in the PM was examined using a fast and highly sensitive magnetic remanence measurement. The longest data set studied was 160 days, significantly longer than that of similar magnetic PM studies improving the statistical robustness. It was found that the variations in the massdependent magnetic parameters displayed a complicated relationship governed by both the meteorological conditions and the PM loading rate, whereas mineralogy/grain-size-dependent magnetic parameters displayed little variation. A six-fold increase in the number of vehicles passing the sampling locations only doubled the magnetic remanence of the samples, suggesting that the measured magnetic signature is in addition strongly influenced by dispersion rates. At both localities the saturation isothermal remanent magnetisation (SIRM) was found to be strongly correlated with the PM mass, and it is suggested that measuring SIRM as a proxy for PM monitoring is a viable alternative to magnetic susceptibility when the samples are magnetically too weak. The signal was found to be dominated by magnetite-like grains less than 100 nm in diameter which is thought to be derived primarily from vehicles. Such small grains are known to be particularly dangerous to humans. There was also evidence to suggest from magnetic stability parameters that the magnetite-like grains were covered with an oxidised rim. The concentration of magnetic PM was in the range of 0.3– 0.5% by mass. r 2003 Elsevier Ltd. All rights reserved. Keywords: Magnetic measurements; Iron oxide; Vehicle pollution; Urban PM
1. Introduction Urban atmospheric particles are known to adversely effect human health (Harrison and Yin, 2000; Zhu et al., 2001). Hence, there is a need to identify both the constituents and atmospheric transport pathways of urban particulate matter (PM). Due to the high content of magnetic minerals in urban PM, the application of *Corresponding author. Fax: +44-131-668-3184. E-mail address: [email protected] (A.R. Muxworthy).
magnetic techniques to rapidly assess urban PM is increasingly being seen as a viable approach (e.g., Matzka and Maher, 1999; Shu et al., 2000, 2001). Magnetic minerals are common amongst pollution PM, with bulk iron content found to constitute 1% of urban atmospheric PM (Department of the Environment, 1996), and iron oxides and hydroxides typically contributing to 10–70% of the bulk iron content (Dedik et al., 1992; Weber et al., 2000). Iron impurities in fossil fuels convert on combustion to magnetic iron oxides, i.e., magnetite, hematite or a mixture of both. Due to its combustion origin, magnetic PM is not only dangerous
1352-2310/03/$ - see front matter r 2003 Elsevier Ltd. All rights reserved. doi:10.1016/S1352-2310(03)00500-4
ARTICLE IN PRESS 4164
A.R. Muxworthy et al. / Atmospheric Environment 37 (2003) 4163–4169
in itself (Phumala et al., 1999; Garcon et al., 2000, 2001), but is also associated with other hazardous pollutants which are injected into the atmosphere during combustion (Morris et al., 1995; Muxworthy et al., 2001). PM in vehicle exhaust is dominated by particles o 10 mm, i.e., PM10 (Kim et al., 2001). In addition to direct combustion derived particles in urban environments vehicles produce other magnetic PM via abrasion/corrosion in . particular brake wear (Olson and Skogerboe, 1975; Osterle et al., 2001), which also contains particles o 10 mm (McCrone and Delly, 1973; Gillies and Gertler, 2000). The relationship between the magnetic signature of PM and other pollution data, e.g., PM mass and gaseous pollution products, is still not fully understood. There have been a limited number of studies which have tried to resolve this relationship by examining time series (Morris et al., 1995; Muxworthy et al., 2001). These studies have found that the magnetic signature displays a complicated behaviour, which is related to other pollution data, but is also strongly effected by meteorological variations, and in addition is highly site dependent. Morris et al. (1995) found a strong correlation between the magnetic susceptibility of daily sampled PM and pollutants such as NO2. However, susceptibility yields only limited information about the magnetic content if no other magnetic information is known. Muxworthy et al. (2001) extended the work of Morris et al. (1995) by measuring magnetic hysteresis of urban PM. Magnetic hysteresis data is more descriptive than susceptibility data as it provides more information about mineralogy, grain size and concentration (Dunlop . and Ozdemir, 1997), which in turn provides more information about the source and characteristics of the PM. Muxworthy et al. (2001) found strong correlations between grain-size-dependent magnetic parameters and the relative humidity. However in their study, due to the insensitivity of the magnetometer used to measure the magnetic hysteresis, PM samples were measured in weekly batches. This averaging removed much of the signal variation and decreased the quality of the data. In addition, due to the rather long measuring time of hysteresis it was not possible to measure long time series. The initial proposal for this study was to advance and improve on the studies of Morris et al. (1995) and Muxworthy et al. (2001), by measuring time series of PM using magnetic techniques which allowed for: daily samples to be measured, changes in magnetic stability (controlled by grain-size and mineralogy) and total magnetic content to be determined, and for long time series to be measured to reduce statistical error. In Morris et al. (1995) and in particularly in Muxworthy et al. (2001) the number of data points used in the statistical analysis was rather low; as a rough guideline the number of data points should be at least B100. To meet these requirements magnetic remanence mea-
surements were made. Although remanences have significantly smaller magnetic signals than in-field measurements, e.g., magnetic hysteresis or magnetic susceptibility, it is possible to measure remanences by using superconducting quantum interference device magnetometers which are highly sensitive and allow for smaller samples to be measured. A set of experiments were designed which combined remanence inducement and partial demagnetisation to assess magnetic stability, and had a relatively short measurement time.
2. Samples and methodology PM samples were collected by the Bayerisches Landesamt fur . Umweltschutz (BLFU, Bavarian Department for the Environment) at the permanent air monitoring stations at Luise-Kiesselbach-Platz (LKP, EU code DEBY085) and WestendstraXe (WS, EU code DEBY045). PM10 was collected using Eberline FH 62I-N dust sampler over 24-h periods. At WS the filter samples were measured between 1 April 2000–7 September 2000 (160 days) and at LKP 24 August 2000–31 October 2000 (69 days). PM10 concentration was measured by its absorption of beta rays and the sampled air volume per time unit was kept constant, taking into account air temperature and pressure. The WS station is 4 km west of the city centre and classified as being in the city (Fig. 1), while LKP is 3 km south of WS and is classified as being in the city suburbs (BLFU, 1998). The prevailing wind direction for Munich is westerly. The daily average number of cars for the period 1990–1995 was 20,000 per day at WS and
48˚16′38
48˚14′08
N DWD
48˚11′38
WS 48˚08′58
LKW
CENTRE
48˚06′28
48˚03′58
PREVAILING WIND DIRECTION
5 km
48˚01′28 11˚25′01 11˚27′31 11˚30′01 11˚32′31 11˚35′01 11˚37′31 11˚40′31
Fig. 1. Map of Munich showing the locations of the two pollution monitoring sites LKW (DEBY085) and WS (DEBY045), and the Deutscher Wetter Dienst (DWD, German Weather Service).
ARTICLE IN PRESS A.R. Muxworthy et al. / Atmospheric Environment 37 (2003) 4163–4169
Table 1 Summary of the daily mean MASS, SIRM and SIRM/MASS for the two data sets in this examined study WS
MASS (mg) SIRM ( 10 9 A m2) SIRM/MASS (mA m2/kg)
LKP
Mean
s
Mean
s
0.6 18.9 32.1
0.2 0.2 0.1
0.8 36.5 44.3
0.2 0.4 0.1
The daily means were calculated using a log-normal distributions, and converted back to a linear scale. s is the standard deviation of the normal distribution, and is not symmetric about the mean on a linear scale. At WS the number of measurements used to calculate s was 160, and at LKP 69.
SIRM
normalised remanence
117,000 per day at LKP (BLFU, 1998). The WS sampling station is situated 6 m from a road, whilst the LKP station is located 2 m from the road. At WS there is a tram line running along the road. Both samplers are placed 4 m above the ground. In addition to being located close to a road, WS is located approximately 50 m from a workshop for public transport buses and trams, which could potentially contribute magnetic PM not necessarily associated with gaseous pollution products. At both sites in addition to PM10 concentration, CO, NO, NOx (NO+NO2) and SO2 were measured. The sampling techniques and instrumentation for this data is fully described in BLFU (1998). Meteorological data was provided by the Deutscher Wetter Dienst (DWD, German Weather Service). Air temperature, relative humidity, wind speed and rainfall were measured at the meteorological station ‘‘Munchen . Stadt’’ of the DWD, which is located B4 km NNE from WS (Fig. 1). Wind direction data was not considered, because of the effects of localised wind channelling. This channelling effect may also effect the quality of the wind speed data. The dust filters provided by the BLFU were 4 cm broad bands containing series of circular dust spots with 2 cm diameter. Each spot contains the dust sampled from 24 m3 of air during a 24 h sampling interval starting at 0:00 local time. Razor blades were used to cut out the dust covered filter material. These circular filter pieces were folded several times to contain the dust inside the filter material and firmly fixed in a 1 cm long piece of plastic drinking straw. This sample preparation allowed manipulation without loosing any of the dust and provided the necessary mechanical stability for magnetic dust particles to carry a stable remanence direction. The samples were given a saturation isothermal remanent magnetisation (SIRM) in a field of 1 T with an electromagnet. From hysteresis measurements on similar PM10 samples collected in Munich, such a field is known to be strong enough to induce a SIRM (Matzka, 1997; Muxworthy et al., 2001). The samples were demagnetised using a 2 G alternating field (AF) demagnetiser. After each treatment, the straw was measured in two positions with a cryogenic magnetometer in a fieldfree room. To test the reliability of this method, the remanence carried by empty drinking straws and blank filter paper were measured. The SIRM of the sample holder plus the empty straw gives a magnetic moment of B6 10 11 A m2 close to the detection limit of the cryogenic magnetometer. The blank filter paper gives values close to 1.4 10 9 A m2, some 13 times smaller than the signal from average SIRM signal of WS (19 10 9 A m2) and 25 times smaller than the signal from LKP (36 10 9 A m2) (Table 1). In general the signal of the filter paper was small compared to that of the dust, and was considered negligible in this study.
4165
1.0
AF10
0.8
37 nm
AF25
0.6 76 nm
0.4
215 nm
0.2 0.0 0
10
20
30 40 AF field (mT)
50
60
Fig. 2. AF demagnetisation curve for a test sample from WS (24 August 2000). Also shown are the AF demagnetisation curves for three synthetic magnetite samples with mean grain sizes of 37, 76 and 215 nm, from Dunlop and Argyle (1991).
To measure a sufficient number of samples for statistical analysis, a fast measuring scheme was adopted. Samples were produced from the filter bands, magnetically treated and measured in batches of eight. We reduced the information content of each AF demagnetisation curve to two representative parameters; firstly a SIRM, which has been found to be representative of the total magnetic content of the sample. Secondly the ratio AF25/AF10 which is the ratio of SIRM AF demagnetised with a 25 mT field divided by SIRM AF demagnetised with a 10 mT field (see Fig. 2). This ratio assesses the magnetic stability of the sample against AF demagnetisation. If the magnetic signal is dominated by one mineral, then AF25/AF10 is controlled only by magnetic domain state which is directly related . to grain size or internal structure (Dunlop and Ozdemir, 1997). We considered this ratio instead of AF25/SIRM, because we wanted to assess variations in only higher coercive fractions of remanence. Thermomagnetic curves measured by Muxworthy et al. (2001) found that these samples had only one significant phase which was magnetite-like.
ARTICLE IN PRESS A.R. Muxworthy et al. / Atmospheric Environment 37 (2003) 4163–4169
4166
that the PM10 has, like NO, a short residency time in the air.
3. Results 3.1. Magnetic measurements
3.2. Daily measurements The normalised day-of-week averages for the magnetic and pollution data for WS and LKP are plotted in Fig. 3. All the data sets at WS and LKP, with the exception of AF25/AF10, display low-weekend values (Sunday lower than Saturday). The Thursday low in LKP data set is attributed to the very high average rainfall on Thursdays, which occurred during the relatively short time series (69 days). The PM data, i.e., PM10 and SIRM, appear to be particularly affected by the precipitation. AF25/AF10 displays virtually no week–day variation, suggesting only one magnetic source. That the PM10 and SIRM data display very similar daily behaviour to the NO data, suggests that the magnetic signature has both the same source as NO and
3.3. Principal component factor analysis The magnetic results were compared to the pollution and meteorological data using principal component factor analysis (PCFA). The magnetic variables and PM10 were transformed using logarithmic base 10 function in a similar manner to that described elsewhere in the literature (e.g., Morris et al., 1995; Urbat et al., 1999). This transformation accommodates the lognormal distribution of these parameters (Kim et al., 2001) . The meteorological data (precipitation, temperature, wind speed and relative humidity) were not measured at the same locations at the pollution data; however, they were taken as global, that is, there would be little spatial variation within Munich in the variance when averaged daily. The variable most likely not to be global being the precipitation. To simplify and reduce the matrix during PCFA, only representative meteorological, pollution and magnetic data was selected. The meteorological data was represented by precipitation, temperature, wind speed and relative humidity, the pollution data by CO, NO and PM10. Both NOx and SO2 were strongly correlated with both CO and NO. The magnetic data was represented by SIRM and AF25/AF10. Bi-plots of some combinations of these parameters are shown in Fig. 4.
1.5
1.0
1.0
0.5
0.5
NO
SIRM
1.5
SIRM (×10-7Am2)
1
+ ++ + + + + + ++ ++ + +++++++++++++ +++ ++++++++++ ++ +++ ++ + + + + + + + ++ ++++ ++++++++++ + + +++ + +++++ +++ + ++++ +++ +++++++++++ ++ +++ ++ ++ + ++++ ++ ++++ + + + ++ + + + + ++
0.1 +
0.1
1
(a) 0.0
+ + ++ + ++ + + + + ++ ++ + + + + + +++ ++ + ++ +++++++++ +++++++ ++++ ++++ + + ++++ + + + + + ++ ++++++++ + + ++ + + + ++++ ++++++++++ + ++ + ++ + + ++ + + ++ ++ ++++++++ + +++ +++ + + + + ++++ + + + ++ + + +++ + +
0.1
+
0.0
(b)
MASS (mg)
0.5
1.0
1.5
CO (mgm-3)
0.0
(c)
0.8 1
1.5
PM 10
1.0
0.5
0.5
0.0
0.0
SIRM (×10-7Am2)
1.0
AF25 /AF10
1.5
AF25 /AF10
(a)
1
+ WS LKP
SIRM (×10-7Am2)
Complete AF demagnetisation curves were measured for a representative selection of samples. In Fig. 2 a typical AF demagnetisation curve is shown together with standard curves for synthetic, sized magnetite samples. Several features can be seen, firstly at lowfields the curve for the PM10 sample lies near that of the 76 nm magnetite sample. As the field increases the PM10 sample’s curve decreases less rapidly than those of the synthetic magnetite samples, and is magnetically more stable.
0.7 +
0.6
+ + ++++++ + + +++++ ++++ + +++++ +++++ ++++ ++ ++++ ++++ ++++++++++++ ++++++++++++++ ++ + ++ + + +++++++ + +++ + ++++ +++ +++ +++++ ++++ + ++ + ++ ++++ ++++ + + + + +
+ +++ ++ + + ++++ ++ + +++ + ++ + ++ +++++++++ + ++ ++++ ++ ++++++++++++ +++ +++ ++ + + +++ + +++++++ ++ ++ + + ++ + + + ++ ++++++++++++++++ ++ + ++++ ++ +++ +++++ +++ ++ +++ + ++ ++++
0.1
+
0.5
(b)
Sa Su M T W Th F week day
(d)
Sa Su M T W Th F week day
Fig. 3. Normalised day-of-week averages from LKP (light grey) and WS (dark grey) for (a) CO, (b) PM10, (c) SIRM and (d) the ratio AF25/AF10. The parameters are normalised by the total average value.
(c)
0.1 SIRM (×10-7Am2)
1
0
(d)
2 4 wind speed (ms-1)
Fig. 4. Bi-plots of various parameters plotted against each for data from WS and LKP: (a) SIRM versus PM10; (b) SIRM against CO; (c) AF25/AF10 against SIRM; and (d) SIRM versus wind speed. SIRM versus NO (not shown) displays similar trends to (b).
ARTICLE IN PRESS A.R. Muxworthy et al. / Atmospheric Environment 37 (2003) 4163–4169 1.0 CO
WS
NO
0.5
factor 2
RAIN
PM 10
RH 0.0 AF25 /AF10 SPEED
TEMP
SIRM
-0.5
-1.0
(a) 1.0 CO
LKP
NO
PM10
SIRM
0.5
factor 2
RH
AF25 /AF10
0.0 TEMP RAIN -0.5 SPEED
-1.0 -1.0
(b)
-0.5
0.0
0.5
1.0
factor 1
Fig. 5. Factor loading plots derived from PCFA for the magnetic data (SIRM and AF25/AF10) sequences from (a) WS and (b) LKP, with representative pollution and meteorological data. MASS is the log10 of the mass of the samples, RAIN is the precipitation, SPEED is the wind speed, RH is the relative humidity and the pollutants CO and NO.
PM10 and SIRM display a strong linear dependency with LKP having higher mean values than WS (Table 1 and Fig. 4a). Extrapolating the linear regression on to the axes, the trend passed within error through the origin for both LKP and WS, implying that there was no non-magnetic source contributing to the PM10 signal at either locality. The relationship between SIRM and the gaseous pollution products, e.g., CO etc., is less transparent (Fig. 4b); the data from WS is scattered, whereas LKP displays a stronger linear relationship. There is little variation in AF25/AF10 as SIRM increases (Fig. 4c), suggesting a consistent loading material. The relationship between SIRM and the meteorological data is less clear (Figs. 4d). Increases in wind speed reduce SIRM, suggesting that the source of SIRM is local. This effect is more pronounced in the LKP data. The factor plot of the pollution, meteorological and magnetic data for WS is shown in component space for the first two components in Fig. 5a. The first two factors contribute only 56% of the variance, while the third
4167
component contributes a further 13% and the fourth 11%, however, the main relationships are still visible. PM10 and SIRM plot close to each other as they are strongly correlated with a Spearman’s r significant at the 0.01 level. These two variables are correlated with temperature which is anti-correlated with the precipitation, relative humidity and wind speed. The pollution data represented by CO and NO display a strong correlation with each other, but display no strong correlation with any other variable. The lack of a strong relationship is also seen in Fig. 4b. AF25/AF10 is displays little variation, and plots between the origin and the temperature. In Fig. 5b, the pollution, meteorological and magnetic data for LKP are plotted in component space for the first two factors representing 60% of the variance. The third factor contributed only 14% of the variance. On the factor 1 axis the meteorological data, with the exception of precipitation, plot in similar position as in Fig. 5a, and CO and NO plot on the factor 2 axis. The wind speed data is shifted slightly and lies closer to the factor 2 axis. The sampling period for the LKP data is different to that of WS. PM10 and SIRM are still closely correlated with each other (ro0:01), however, both parameters now lie closer to CO and NO. There is a stronger correlation with the gaseous pollution data for SIRM (ro0:01) than at WS. The wind speed data is strongly anti-correlated with PM10 and SIRM (ro0:01).
4. Discussion For low-fields the AF demagnetisation curves suggest that the magnetite-like grains are o100 nm, however, as the AF increases the PM10 samples display a higher magnetic stability than for the synthetic samples (Fig. 2). The high stability may indicate the presence of a second phase. This phase is most likely be to an oxidised surface layer; a maghemite rim or near-maghemite rim on the magnetite particles. Due to differences in lattice spacing between magnetite and maghemite, such rims induce a surface stress which effectively increase the stability of the particles giving rise to underestimates for the mean magnetic grain size. Such oxidation processes occur initially very rapidly, however, once the rim is formed further oxidation of the magnetite core occurs very slowly. This may also explain the difference between the mean grain size found in this study compared to previous studies for PM10 from Munich which found a magnetite-like phase with a mean grain size 200–500 nm (Muxworthy et al., 2001). There are clear differences between the two locations studied. The correlation between the magnetic parameters, e.g., SIRM, and the gaseous pollution products, e.g., CO and NO, is site dependent. At LKP SIRM and PM10 were closely related to CO and NO (Figs. 4b and 5b), which can
ARTICLE IN PRESS 4168
A.R. Muxworthy et al. / Atmospheric Environment 37 (2003) 4163–4169
also be seen in the averaged daily trends (Fig. 3). At WS the correlation between these parameters was lower. Considering that the nearly six times as many cars pass LKP than WS, it is seen that the effect of increased traffic loading produces a closer correlation between pollution and magnetic data, i.e., the relative influence of the meteorological effects decreases as the pollution loading increases. However, the exact relationship is complicated and non-linear; attempts to produce a general non-sitespecific linear regression model were unsuccessful. Are the differences between the mean values for SIRM and PM10 from the two localities (Table 1) due to spatial or temporal reasons? To assess this, we consider the ratios of the mean values for the data from the 15 days where the two time series overlap. The ratios for LKP over WS for SIRM and PM10, are respectively B2 and B1.4, implying that the differences are due to spatial variation. It is possible to determine the magnetic content as a percentage by mass if we assume two things; firstly that magnetite is the primary magnetic mineral and secondly that the saturation magnetisation (MS ) can be determined by taking the mean SIRM=MS ratio B0.11 from Muxworthy et al. (2001). By extrapolating from SIRM to determine MS ; we are also incorporating the nonremanence carrying grains, i.e., superparamagnetic grains. For WS the percentage by mass of magnetite is determined to be 0.32% using the data in Table 1, which is identical to that reported for PM70 in Muxworthy et al. (2001), and 0.44% for LKP. Again it is seen that a sixfold increase in vehicles produces only a relatively small increase in the magnetic content of the sample, implying that either vehicles are not large contributors to the magnetic signal, or more likely magnetic PM10 distributes quickly to give an average background level. Matzka and Maher (1999), Hoffmann et al. (1999) and Moreno et al. (2003) found that roadside pollution was mostly deposited within the first few metres from the road suggesting that the pollution is not evenly distributed. Muxworthy et al. (2002) found no significant difference between PM collected at 15 and B150 m from a road in central Munich. Of interest to palaeoclimate studies is how the magnetic signatures are effected by meteorological conditions. Increases in wind speed, precipitation and relative humidity all decrease PM10 and SIRM, though for different reasons. Increases in wind speed are likely to move PM to other areas away from the source station. Relative humidity and precipitation act differently by reducing residency time (Muxworthy et al., 2001). That is, precipitation and relative humidity increase local deposition rates, whereas increasing wind speed decreases it. The reduction of PM10 and SIRM with increasing relative humidity is attributed to hygroscope interaction between SO2 absorbed on the surface of magnetite
grains with water vapour within the air as discussed by Muxworthy et al. (2001). However, Muxworthy et al. (2001) noticed a subtler interaction between the magnetic grain size distribution and the relative humidity; from hysteresis measurements, they found that the smallest grains were preferentially removed by increases in relative humidity. No such relationship was found between AF25/AF10 and the relative humidity in this study. It is suggested that the AF25/AF10 is less sensitive than hysteresis measurements at determining changes in grain size and mineralogy. The relationship between SIRM and wind speed (Fig. 4d) is non-linear. There appears to be a cut-off wind speed above which it is not possible to obtain higher SIRM values at these two locations. It is uncertain if this is a real effect or a technical sampling error. If a real effect, it is likely that the critical wind speed is dependent on the loading rate and location.
5. Conclusions Detailed analysis of the behaviour of the magnetic signature of daily sampled suspended particulate matter (PM) collected in Munich in 2000 reveals that the magnetic signature displays a complicated non-linear relationship with both gaseous pollutants and the meteorological data. The saturation magnetic remanence was found to be strongly correlated with PM10 suggesting that the behaviour of PM10 is adequately described by the behaviour of SIRM. This is of course highly site dependent and assumes that PM loading is from a single homogenous magnetic source. Muxworthy et al. (2001) found that at another location in Munich (Pasing) that this relationship did not hold. However, at these two localities it is thought that there is only one consistent magnetic source, i.e., no non-magnetic sources, and that this is most likely to be derived from vehicles. The magnetic signature appears to be dominated by a magnetite-like grains o100 nm in diameter, probably with an oxidised rim. Such small grains are known to be particularly dangerous to humans as they can be inhaled deep into the lungs (Smith and Aust, 1997; Donaldson et al., 1998). At the two locations there was found to be little variation in the mineralogy and/or grain size of the magnetic PM. Measuring remanence rather than the more standard magnetic susceptibility, does not greatly increase measurement time, however, because of increased sensitivity it allows for significantly smaller samples to be measured. Moreno et al. (2003) found a strong correlation between magnetic susceptibility and SIRM in their study of PM trapped on leaves. It is suggested therefore that in cases where the samples are too weak to measure the susceptibility, SIRM studies provide a viable alternative.
ARTICLE IN PRESS A.R. Muxworthy et al. / Atmospheric Environment 37 (2003) 4163–4169
Acknowledgements The authors thank Mr. Weber of the Bayerisches Landesamt fur . Umweltschutz for providing both the filter samples and the pollution data. This work was funded by the European Union (Contract no. ERBFMXCT 98-0247), as part of the European Network for Mineral Magnetic Studies of Environmental Problems, and NERC grant NER/A/S/2001/00539.
References Bayerisches Landesamt fur . Umweltschutz (BLFU), 1998. Lufthygienischer Jahresbeircht 1997. Bayerisches Landesamt fur . Umweltschutz, Munchen. . Dedik, A.N., Hoffmann, P., Ensling, J., 1992. Chemical characterization of iron in atmospheric aerosols. Atmospheric Environment 26A (14), 2545–2548. Department of the Environment, 1996. Airborne particulate matter in the United Kingdom, Third Report of the Air Quality of Urban Air Review Group, Department of the Environment, UK. Donaldson, K., Li, X.Y., MacNee, W., 1998. Ultrafine (nanometre) particle mediated lung injury. Journal of Aerosol Science 29, 553–560. Dunlop, D.J., Argyle, K.S., 1991. Separating multidomain and remanence induced in pseudo-single-domain magnetites (215–540 nm) by low-temperature demagnetization. Journal of Geophysical Research 96 (2), 2007–2017. . . 1997. Rock Magnetism: FundaDunlop, D.J., Ozdemir, O., mentals and Frontiers. Cambridge Studies in Magnetism. Cambridge University Press, Cambridge, 573pp. Garcon, G., Shirali, P., Garry, S., Fontaine, M., Zerimech, F., Martin, A., Hannothiaux, M.H., 2000. Polycyclic aromatic hydrocarbon coated onto Fe2O3 particles: assessment of cellular membrane damage and antioxidant system disruption in human epithelial lung cells (L132) in culture. Toxicology Letters 117 (1–2), 25–35. Garcon, G., Gosset, P., Garry, S., Marez, T., Hannothiaux, M.H., Shirali, P., 2001. Pulmonary induction of proinflammatory mediators following the rat exposure to benzo(a)pyrene-coated onto Fe2O3 particles. Toxicology Letters 121 (2), 107–117. Gillies, J.A., Gertler, A.W., 2000. Comparison and evaluation of chemically speciated mobile source PM2.5 particulate matter profiles. Journal of the Air and Waste Management Association 50 (8), 1459–1480. Harrison, R.M., Yin, J.X., 2000. Particulate matter in the atmosphere: which particle properties are important for its effects on health? Science of the Total Environment 249 (1–3), 85–101. Hoffmann, V., Knab, M., Appel, E., 1999. Magnetic susceptibility mapping of roadside pollution. Journal of Geochemical Exploration 66, 313–326. Kim, W.-K., Kim, S.H., Lee, D.W., Lee, S., Lim, C.S., Ryu, J.H., 2001. Size analysis of automobile soot particles using field-flow fractionation. Environmental Science and Technology 35, 1005–1012.
4169
Matzka, J., 1997. Magetische, elektronenmikroskopische und lichtmikroskopische Untersuchungen an St.auben und Aschen sowie an einzelnen Aschepartikeln. Universit.at M.unchen. Matzka, J., Maher, B.A., 1999. Magnetic biomonitoring of roadside tree leaves: identification of spatial and temporal variations in vehicle- derived particulates. Atmospheric Environment 33, 4565–4569. McCrone, W.C., Delly, J.G., 1973. The Particle Atlas, 2nd Edition. Ann Arbor Science Publishers, Ann Arbor. Moreno, E., Sagnotti, L., Dinars-Turell, J., Winkler, A., Cascella, A., 2003. Biomonitoring of traffic air pollution in rome using magnetic properties of tree leaves. Atmospheric Environment 37 (21), 2967–2977. Morris, W.A., Versteeg, J.K., Bryant, D.W., Legzdins, A.E., McCarry, B.E., Marvin, C.H., 1995. Preliminary comparisons between mutagenicity and magnetic-susceptibility of respirable airborne particulate. Atmospheric Environment 29, 3441–3450. Muxworthy, A.R., Matzka, J., Petersen, N., 2001. Comparison of magnetic parameters of urban atmospheric particulate matter with pollution and meteorological data. Atmospheric Environment 35 (26), 4379–4386. Muxworthy, A.R., Schmidbauer, E., Petersen, N., 2002. Magnetic properties and M.ossbauer spectra of urban atmospheric particulate matter: a case study from Munich, Germany. Geophysical Journal International 150 (2), 558–570. Olson, K.W., Skogerboe, R.K., 1975. Identification of soil lead compounds from automotive sources. Environmental Science and Technology 9, 227–230. . Osterle, W., Griepentrog, M., Gross, Th., Urban, I., 2001. Chemical and microstructural changes induced by friction and wear of brakes. Wear 251, 1469–1476. Phumala, N., Ide, T., Utsumi, H., 1999. Non-invasive evaluation of in vivo free radical reactions catalyzed by iron using in vivo ESR spectroscopy. Free Radical Biology and Medicine 26 (9–10), 1209–1217. Shu, J., Dearing, J.A., Morse, A.P., Yu, L.Z., Li, C.Y., 2000. Magnetic properties of daily sampled total suspended particulates in Shanghai. Environmental Science and Technology 34 (12), 2393–2400. Shu, J., Dearing, J.A., Morse, A.P., Yu, L.Z., Yuan, N., 2001. Determining the sources of atmospheric particles in Shanghai, China, from magnetic and geochemical properties. Atmospheric Environment 35 (15), 2615–2625. Smith, K.R., Aust, A.E., 1997. Mobilization of iron from urban particulates leads to generation of reactive oxygen species in vitro and induction of ferritin synthesis in human lung epithelial cells. Chemical Research in Toxicology 10, 828–834. Urbat, M., Dekkers, M.J., Vriend, S.P., 1999. The isolation of diagenic groups in marine sediments using fuzzy c-means cluster analysis. In: Tarling, D.H., Turner, P. (Eds.), Palaeomagnetism and Diagenesis in Sediments, Vol. 151. Geological Society, London, pp. 85–93. Weber, S., Hoffmann, P., Ensling, J., Dedik, A.N., Weinbruch, S., Miehe, G., Gutlich, P., Ortner, H.M., 2000. Characterization of iron compounds from urban and rural aerosol sources. Journal of Aerosol Science 31 (8), 987–997. Zhu, R.X., Shi, C.D., Suchy, V., Zeman, A., Guo, B., Pan, Y.X., 2001. Magnetic properties and paleoclimatic implications of loess–paleosol sequences of Czech Republic. Science in China Series D-Earth Sciences 44 (5), 385–394.