Introduction

Soil contamination or soil pollution poses a serious global threat to our environment. Increased concentration of soil pollutants should be of concern worldwide because of their toxicological impacts on ecosystems. In particular, heavy metals can cause severe adverse effects on any non-targeted organism. It is generally considered that heavy metals accumulated in soils are derived from natural lithogenic or artificial anthropogenic sources (e.g., Karimi et al. 2017). In the absence of heavy industry, roadside dust is undoubtedly the most common sources of environmental pollution in urban areas (e.g., Bućko et al. 2011). Recently, the applicability of magnetic monitoring has increasingly been applied for the evaluation of anthropogenic pollution (e.g., Bućko et al. 2010).

Geochemical investigation provides the highest resolution in environmental studies in terms of quantitative evaluation of the degree of pollution as well as qualitative assessment of the pollution sources (e.g., Jordanova et al. 2013). However, geochemical investigation requires a great deal of chemical analyses, which are time-consuming, expensive, and destructive. In contrast, magnetic methods are rapid, cost effective, and non-destructive (e.g., Evans and Heller 2003). In practice, magnetic techniques provide a descent basis for a comprehensive geochemical investigation. Among various rock magnetic approaches, magnetic susceptibility is the most widely used tool to map and monitor the heavy metal pollution of soils (e.g., Hanesch et al. 2007; Yang et al. 2007, 2009; Blaha et al. 2008; Karimi et al. 2011, 2017; Dankoub et al. 2012; Ayoubi et al. 2019a, b; Ayoubi and Adman 2019; Ayoubi and Karimi 2019). Magnetic susceptibility reflects the amount of paramagnetic and ferromagnetic minerals in rocks and soils (e.g., Evans and Heller 2003).

In general, magnetic minerals present in soils have three distinctively different origins. First, unpolluted soils contain lithogenic magnetic minerals that were the product of weathering and transportation of parent rocks. Second, pedogenic process under aerobic conditions is responsible for the enhancement of magnetic susceptibility, by increasing the concentration of nano-sized magnetite and/or maghemite (e.g., Torrent et al. 2010; Liu et al. 2012; Ayoubi and Karami 2019). Third, polluted soils commonly contain significant amount of anthropogenic magnetic minerals. Presence of highly magnetic particles in polluted soils originates from anthropogenic contribution. Urban airborne particulates are enriched in heavy metals, acting as carriers of potentially hazardous heavy metals (e.g., Boyko et al. 2004). Magnetic properties have been used as a proxy for the heavy metal concentrations in soils (e.g., Hanesch and Scholger 2002; Jelenska et al. 2004; Schmidt et al. 2005; Dong et al. 2014; Grison et al. 2015; Ayoubi and Karimi 2019). Such a linkage came from the fact a combustion process is responsible for the enhancement of magnetic particulates and heavy metals in soils (e.g., Tolonen and Oldfield 1986). Therefore, soil magnetometry is useful as a first-order constraint on the degree of heavy metal pollution (e.g., Bućko et al. 2010; Ayoubi et al. 2019a, b).

In industrial regions, fine-grained magnetic particulate matters are produced by combustion of fossil fuels and other high temperature technogenic processes (e.g., Hanesch et al. 2003; Spiteri et al. 2005; Jordanova et al. 2006; Yan et al. 2011; Zhang et al. 2011; Zhu et al. 2013). On the other hand, the most important source of heavy metal pollutants in urban environments is roadside dust emission (e.g., Muxworthy et al. 2001; Gautam et al. 2004; Lu and Bai 2006; Kim et al. 2007; Lu et al. 2007; Werkenthin et al. 2014; Ojha et al. 2016). For instance, the primary constituent of automobile exhaust was found to be associated with magnetic particles. In addition, exterior rusting, internal exhaustion, and ablation of breaks also produce fine-grained magnetic particulate matters (e.g., Petrovský et al. 2000; Wawer et al. 2015; Ma et al. 2016). Thus, rock magnetic proxies have been applied to traffic-related roadside dusts (e.g., Prajapati et al. 2006; El-Hasan 2008; Maher et al. 2008; Lu et al. 2011; Nikolaeva et al. 2017; Pierwola et al. 2020). However, the correlation between pollutants and magnetic proxy was far from being a simple linear relation (e.g., Chaparro et al. 2004). As a result, it was necessary to generate magnetic mapping at a regional scale to assess the spatial distribution of pollutants for each independent investigation (e.g., Ayoubi and Adman 2019).

A regional scale magnetic investigation on roadside dust is rare in Korea (Kim et al. 2007). In this study, we tested magnetic measurements as a proxy to monitor the degree of roadside pollution. The main objective of this work was to identify detailed characteristics of the magnetic materials derived from circulation of traffic-derived roadside soils in Daejeon, South Korea. To unravel magnetic signal on the roadside dusts, low-field magnetic susceptibility, frequency dependence of magnetic susceptibility, susceptibility of anhysteretic remanent magnetization, isothermal remanent magnetization (IRM) acquisition and demagnetization, back-field IRM treatment, and thermal demagnetization of composite IRM will be carried out. In particular, we focused on the mutual relation between various high-field IRM-related properties.

Methods and materials

Description of the study area and sampling

South Korea is located in a transition zone between the humid subtropical zone and the humid continental climate zone. The monsoon-influenced climate is dominant over the summer and brings about 80% of the annual rainfalls. Daejeon is South Korea's fifth-largest metropolis with a population over 1.5 million (Fig. 1). In 2019, annual precipitation from rainfalls (and snowfalls) in Daejeon was 1458.6 mm. According to climate data from 1981 to 2010, monthly mean temperatures of Daejeon ranged from − 1.0 °C in January to 25.6 °C in August. The mean elevation of Daejeon is 70 m a.s.l. A Jurassic two mica granite constitutes the basement of Daejeon area (Hwang and Moon 2018).

Fig. 1
figure 1

In situ magnetic susceptibility traverses across urban street in Daejeon, South Korea measured on June 28, 2019. Magnetic susceptibilities were measured from a total of six profiles (N1, N2, N3, S1, S2, and S3). Each profile was 100 m long, set to be nearly perpendicular to the twelve-way street along the river. The inset was generated from the “Generic Mapping Tools (GMT)” version 6.0.0 (released on November 1, 2019) distributed from the School of Ocean and Earth Science and Technology, University of Hawaii at Manoa (https://gmt.soest.hawaii.edu/). We set the base station (star symbol) near to the main street (36°21′32.19″N, 127°21′27.49″E)

Full size image

The meteorological parameters on the sampling date (June 28, 2019) were available from the Korea Meteorological Administration (https://www.weather.go.kr/). In detail, temperature ranged between 22.1 and 29.9 °C, while humidity was 76.3%. The prevailing wind directions on June 28, 2019 were SE and SSE, with low wind speeds of 1.5 m/s.

The Hanbat-daero is a major twelve-way street (Fig. 1), with the daily traffic volume over 100,000 vehicles on weekdays. It is not awkward to witness a heavy traffic queue during rush hours. The stream bridge (Yuseong-daegyo) is a street-stream crossroad, which is about 500 m away from the highway entrance. The deck height of the stream bridge is about 5 m. At ground level of the bridge, there is a grassland, where we expect to check spatial variation of traffic-related pollution (Fig. 1). We set the base station (star symbol in Fig. 1) near to the main street (36°21′32.19″N, 127°21′27.49″E). It is likely that vehicles are the only significant source of pollution in the grassland, because there is no nearby agricultural or industrial facilities (Fig. 1).

Magnetic susceptibilities were measured from a total of six profiles, three towards the south and three towards the north (Fig. 1). Each profile was 100 m long along the river, set to be nearly perpendicular to the twelve-way street (Fig. 1). Three profiles towards the south were separated by 5 m with one another, labeled as S1, S2, and S3. Similarly, three profiles towards the north were labeled as N1, N2, and N3. The net investigation area covered about 2000 m2 (200 m long and 10 m wide).

Magnetic susceptibilities were measured using a portable magnetic susceptibility meter ZH-instruments SM30 with a sensitivity of 10–7 SI. A total of 238 readings were taken (40 from S1, 31 from S2, 32 from S3, 40 from N1, 54 from N2, 41 from N3). Each data point was double or triple checked to ensure the reproducibility of signal. A total of 76 representative sites were selected (13 from S1, 10 from S2, 10 from S3, 13 from N1, 18 from N2, 12 from N3), where we collected samples from the upper 5 cm of the topsoils. These 76 sites were picked on the basis of magnetic susceptibility measurements including sites from the highest, lowest, and median (or mean if necessary) susceptibilities. The topsoils were un-stratified with loose, brown fine- to medium grains sand with occasional fine gravels. Whenever encountered, we discarded gravels. Samples were collected using a stainless steel trowel, and stored in a non-magnetic one-inch cube. Collected soil samples were placed in a sealed polyethylene bags and transported to the laboratory for further measurements. For these plastic cubes, magnetic susceptibilities were measured in the lab using a Bartington MS2 susceptibility bridge.

Presence of void induces grain motions in a plastic cube, resulting in inconsistent magnetic behavior. To prevent possible grain motions during magnetic measurements on spinning magnetometers, non-magnetic paraffin was used. It is well-known that non-magnetic solid paraffin has a melting point of ≈40 °C. We injected liquid paraffin into empty spaces of plastic cubes and dried overnight. The remanent magnetization of a one-inch plastic cube with full paraffin had remanent magnetization 3000 fold lower than the weakest soil samples. Hence, the influence of paraffin injection was magnetically insignificant in the present study. Samples were weighed, and then stored in a refrigerator to minimize the effect of drying or moisturizing.

Laboratory analyses

Magnetic susceptibilities (χ) were measured at low (0.46 kHz; χlf) and high frequencies (4.6 kHz, χhf) using a Bartington MS2 dual frequency sensor. The frequency dependence (χfd) was calculated as χfd (%) = [(χlf − χhf)/χlf] × 100 to address the relative contribution of the superparamagnetic (SP) particles in the total magnetic fraction of the soils (Dearing et al. 1996).

Anhysteretic remanent magnetization (ARM) is produced when a steady external field is superimposed on a decaying alternating-field (AF). The susceptibility of ARM was determined as χARM = ARM/BARM for a field BARM of 0.05 mT or 500/4π = 39.8 A/m (Table 1).

Table 1 Magnetic Properties of 26 Soil Samples
Full size table

To eliminate the influence of initial state dependence of remanent magnetization, samples were AF demagnetized to 100 mT before each isothermal remanent magnetization (IRM) treatment. All the remanent magnetizations were determined using an AGICO JR-6 spinner magnetometer. IRM was acquired along z at 18 steps of 1, 2, 5, 10, 20, 25, 30, 40, 50, 60, 80, 100, 120, 150, 200, 300, 500, 1000 mT using an ASC-Scientific IM-10–30 impulse magnetizer. AF demagnetization was carried out at 15 steps of 1.0, 2.5, 5.0, 7.5, 10.0, 12.5, 15.0, 20.0, 25.0, 30.0, 40.0, 45.0, 50.0, 60.0, and 80.0 mT using an AGICO LDA5 AF demagnetizer. On completion of AF demagnetization of IRM, IRM was produced again by exposing samples in a field of 1 T. Then, back-field IRM was applied along negative z at 18 steps of 1, 2, 5, 10, 20, 25, 30, 40, 50, 60, 80, 100, 120, 150, 200, 300, 500, 1000 mT. Note that IRM was produced along positive z, while back-field IRM was applied along negative z, to avoid the possible effect of remanence anisotropy.

All the samples showed saturation by 300 mT, hence the IRM applied at 1 T can be treated as saturation IRM (SIRM). Classical backward S-ratio compares the SIRM acquired in a saturation field (e.g., 1 T) with the IRM in a smaller back-field (e.g., − 300 mT) as S = IRM–0.3 T/SIRM (Stober and Thompson 1979). Revised backward S-ratio was modified as S− 0.1 T = 0.5 (1 − IRM− 0.1 T/SIRM) and S− 0.3 T = 0.5 (1 − IRM− 0.3 T /SIRM) (Bloemendal et al. 1992; Liu et al. 2007). While S varies from − 1 to + 1, S− 0.1 T or S− 0.3 T varies from 0 to + 1. In particular, it has been widely used that values of S− 0.3 T in percent represent the amount of low coercivity fraction in samples. The forward S-ratio compares the SIRM acquired in a saturation field (i.e., 1 T) with the IRM in a smaller forward field (e.g., 100 mT) as S0.1 T = IRM0.1 T/SIRM (Kruiver and Passier 2001). A complementary statistical treatment and physical rationale on the use of S-ratio are beyond the scope of this study, but were provided by Liu et al. (2007) and by Heslop (2009), respectively. In addition to the various forms of S ratio, two other coercivity-related parameters were useful (e.g., Bloemendal et al. 1992; Yamazaki et al. 2003). First, high-coercivity fraction of IRM (HIRM) was defined as HIRM = 0.5 (SIRM + IRM− 0.3 T). Second, medium-coercivity fraction of IRM (MIRM) was defined as MIRM = 0.5 (SIRM + IRM− 0.1 T) − HIRM.

On completion of all the field-treated experiments, representative cubes were dismantled for further analysis. Magnetic particles were separated using a hand magnet. Half of the extracted magnetic particles were examined under the microscope. Plastered caps were employed for the remaining half of the extracted magnetic particles. These caps were then one-inch cut for thermal treatment. At last, stepwise thermal demagnetization of 3-axis IRM (Lowrie 1990) was applied to the representative samples on the basis of their IRM behavior. IRMs were sequentially produced along the three mutually perpendicular directions using an ASC-Scientific IM-10-30 impulse magnetizer. Sequential fields used to produce a composite IRM were 1.2, 0.4, and 0.12 T along z-axis, x-axis, and y-axis, respectively.

Statistical analysis

Relationship between magnetic parameters in the topsoil was studied by means of Pearson correlation analyses. The Pearson correlation coefficient (R) was used to quantify the degree of linear relation in a bivariate system (Table 2).

Table 2 Pearson correlation coefficients between magnetic parameters (N = 26)
Full size table

Table 2 also lists correlation coefficient of determination (R2) between pairs of twelve rock magnetic factors including χfd, χARM, MAF, Bcr, forward S ratio at 0.1 T (S0.1 T), backward S ratio at − 0.1 T (S− 0.1 T) and at − 0.3 T (S− 0.3 T), HIRM, MIRM, MDF, crossover field (BR), and crossover point (P). Although crossover point has been abbreviated as R in rock magnetism and environmental magnetism communities, we used P to avoid confusion with the Pearson correlation coefficient (R).

Results

Spatial distribution of magnetic susceptibility

The magnetic susceptibility values ranged from 8.6 to 71.7 × 10–5 SI in the soils of profile S1 (Fig. 2). The mean and median magnetic susceptibility of S1 were 29.2 ± 13.1 × 10–5 SI and 28.6 × 10–5 SI, respectively. Two highest peaks occurred at 11.6 and 14.5 m away from the street (Fig. 2). Magnetic susceptibility of profiles S2 decayed to lower value from 0 to 55 m towards the south and then slightly enhanced from 55 to 100 m (Fig. 2). The mean and median magnetic susceptibility of S2 were 25.4 ± 8.6 × 10–5 SI and 21.4 × 10–5 SI, respectively. The profile S3 showed a peak at 62.5 m towards the south (Fig. 2). The mean and median magnetic susceptibility of S3 were 25.2 ± 10.4 × 10–5 SI and 19.5 × 10–5 SI, respectively. The coefficient of variation (CV) or relative standard deviation (RSD) is a percentage ratio of the standard deviation to the mean. The CV or RSD for S1, S2, S3 were 45.0, 33.7, 41.4%, respectively.

Fig. 2
figure 2

Profiles of soil-surface magnetic susceptibility. A total of 238 readings were taken (40 from S1, 31 from S2, 32 from S3, 40 from N1, 54 from N2, 41 from N3)

Full size image

Three profiles towards the North appeared to be fairly similar with one another (Fig. 2). They all showed decreasing trend of magnetic susceptibility from 0 to 55 m, and then jumped to higher values (Fig. 2). The magnetic susceptibility values ranged from 13.2 to 53.5 × 10–5 SI in the soils of profile N1 (Fig. 2). The mean and median magnetic susceptibility were 29.9 ± 9.6 × 10–5 SI and 29.0 × 10–5 SI for N1, 27.1 ± 9.8 × 10–5 SI and 25.2 × 10–5 SI for N2, and 30.2 ± 8.1 × 10–5 SI and 28.3 × 10–5 SI for N3, respectively. The CV or RSD for N1, N2, N3 were 31.9, 36.2, 26.7%, respectively. Values of CV or RSD were smaller for northern profiles than those for southern profiles, as the variation of magnetic susceptibility was more substantial in southern profiles (Fig. 2).

On the basis of magnetic susceptibility collected from six profiles, two dimensional map of magnetic susceptibility was constructed (Fig. 3). We found that the 5 and 95% percentile of magnetic susceptibility from a total of 238 readings were 10 × 10–5 SI and 50 × 10–5 SI, respectively. Hence, these values were used as the lower and upper bound of 20-step contours (Fig. 3). A contour map was produced from Grapher 10.5 with the lowest available data smoothing (i.e., smoothing factor = 1). It is evident that high values of magnetic susceptibility were narrowly concentrated adjacent to the street (Fig. 3). Magnetic susceptibility decreased away from the street to 55 m distance (Fig. 3). Magnetic susceptibility began to increase again, reaching 30 to 40 × 10–5 SI in southern profiles and 30 to 50 × 10–5 SI in northern profiles (Fig. 3).

Fig. 3
figure 3

Contour plot of magnetic susceptibility values in the topsoil. Elevation (i.e., depth) profile along the transect was provided as an inset

Full size image

Grain proxy

The mean of χfd was 2.21 ± 1.27%, with only four samples exceeding 4% (Table 1). The low χfd values (< 5%) indicate that SP grains do not exist in considerable amounts in the topsoil of studied samples (e.g., Dearing et al. 1996; Magiera et al. 2011, 2013). Such low χfd was the characteristics of roadside dusts (e.g., Shilton et al. 2005; Yang et al. 2010; Bourliva et al. 2016). χfd and χlf were approximately inversely correlated (Fig. 4a), consistent with the frequency dependence of magnetic susceptibility observed for ferrimagnetic grains (Dearing et al. 1996).

Fig. 4
figure 4

a Plots of χfd percentage versus χlf. b Bivariate plot of χAMS versus χlf. Reference lines were retrieved from King et al. (1982)

Full size image

Plots of χARM versus χlf (Fig. 4b) were useful to detect changes in the relative grain-size of ferrimagnetic grains in natural materials (King et al. 1982). Observed data fell in the range of 0.1–20 μm magnetite particles (Fig. 4b).

Isothermal remanent magnetization

The IRM acquisition curves (Fig. 5) were rapidly increasing at low fields (< 100 mT). They reached saturation by 300 mT, indicating a predominance of low-coercivity Fe-oxides (Fig. 5a). The median acquisition field (MAF) was defined as the impulse field that reached the 50% of SIRM. Values of MAF varied from 33.7 mT (GCD31) to 98.9 mT (GCB29), with a mean of 56.2 ± 14.2 mT (Table 1).

Fig. 5
figure 5

a Normalized curves for the acquisition of IRM. b Back-field decay of saturation IRM

Full size image

The coercivity of remanence (Bcr) was determined from the back-field IRM curves (Fig. 5b). Values of Bcr varied from 23.7 mT (GCD40) to 60.1 mT (GCB11), with a mean of 33.9 ± 7.5 mT (Table 1). Values of S− 0.1 T ranged from 0.683 (GCB11) to 0.896 (GCE30), with a mean of 0.823 ± 0.049 (Table 1). Values of S− 0.3 T were larger than those of S− 0.1 T, with a mean of 0.961 ± 0.011 (Table 1). The HIRM reflects the amount of hard magnetic phase (hematite or goethite) in mineral mixtures that also contain magnetite (Liu et al. 2007). High S-ratios (S− 0.1 T and S− 0.3 T) and low HIRM indicate a prominence of low-coercivity magnetic particulates (Table 1). The MIRM reflects the amount of medium-coercivity magnetic phase in mineral mixtures. The MIRM varied from 0.087 (GCE30) to 0.306 (GCB11), with a mean of 0.147 ± 0.055 (Table 1).

The median destructive field (MDF) was defined as the AF that reduced the NRM by 50% (Fig. 6a, Table 1). The MDF thus reflects the coercivity spectra of magnetic particles. Values of MDF ranged from 5.4 mT (GCF15) to 45.2 mT (GCB11), with a mean of 13.7 ± 7.9 mT (Table 1). Intersection point between IRM acquisition and IRM decay curves lied well below 0.5, indicating some degree of grain interaction (Table 1).

Fig. 6
figure 6

Composite diagrams showing a AF demagnetization of saturation IRM, b difference between normalized AF demagnetization curves between observed MBF and synthesized MBF*. Back-field coercivity spectra were constructed from AF demagnetization of SIRM (MAF) with a relation of MBF* = 1  −  2MAF

Full size image

One set of coercivity spectra (MAF) were constructed from stepwise AF demagnetization of SIRM. Another independent set of coercivity spectra (MBF) can be obtained during back-field IRM decay curves. According to Wohlfarth theory (Wohlfarth 1958), non-interacting single domain (SD) particles satisfy the relation of MBF* = 1 − 2 MAF. When MBF was plotted against MAF, ideal non-interacting SD grains would yield a slope of − 2 with a y-intercept at + 1 (Henkel 1964). In reality, observed MBF and synthesized MBF* from MAF would be different unless samples are free of interaction. The difference between observed MBF and synthesized MBF* was represented as ∆M (e.g., Mayo et al. 1990, 1991a, b; Speliotis and Lynch 1991). It is common to observe negative ∆M in most field intervals as interaction tends to lower magnetization (Petrovský et al. 1993). Although rare, positive ∆M is possible in some field intervals, where positive interaction induces enhanced magnetization. In the present study, we found negative ∆M from all samples (Fig. 6b). It should be noted that sample GCB29 showed the deepest negative ∆M and the largest MAF, indicative of strongly interacting fine-grained magnetic particles in this sample (Fig. 6b, Table 1).

The maximum unblocking temperatures of composite IRMs (Lowrie 1990) were 580 °C which is the Curie temperature of magnetite, suggesting that the magnetic mineralogy was dominated by magnetite (Fig. 7a). Under the microscope, we found two different types of magnetic material. About two thirds of magnetic particles were angular in shape, with various sizes from 20 to 400 μm (Fig. 7b). The presence of lithogenic magnetite is clearly shown by angular magnetite with octahedral morphology (Fig. 7b). Other particles showed spherical morphologies with relative uniform diameters of about 50 μm or less (Fig. 7b).

Fig. 7
figure 7

a Thermal demagnetization curves of a three-component (hard, medium, soft) IRM (Lowrie 1990), indicating a presence of magnetite. b Representative images of spherical and angular magnetic particulates

Full size image

Discussion

Spatial distribution of magnetic susceptibility

A generally higher level of magnetic susceptibility was observed near the roadside base of profiles (Figs. 2 and 3). Compared to the peripheries of the studied area, magnetic susceptibilities were much more pronounced getting closer to the main street. It is natural to anticipate that magnetic susceptibility decreases with distance from the input source (e.g., Hoffmann et al. 1999). The highest magnetic susceptibility was observed within a distance of a few meters from the main street. Such characteristics can be related to the emission of the traffic pollution produced by vehicular engine emissions (e.g., Zechmeister et al. 2005; Bućko et al. 2010; Werkenthin et al. 2014; Wawer et al. 2015; Ma et al. 2016; Nikolaeva et al. 2017; Pierwola et al. 2020). Magnetic enhancement getting closer to the main street decayed exponentially from 0 to 55 m with respect to the distance (Figs. 2 and 3) due to magnetic grain size or the amount of magnetic particles with distances away from the source.

In the present study, magnetic susceptibility was lower away from the road, because air circulation was intense along the river, less efficient in settling of magnetic particles on the grassland. The lowest values of magnetic susceptibility occurred about 55 m away from the main street and a systematic increase of susceptibility towards the roads was observed (Fig. 2). Perhaps, the air flow became less significant to settle magnetic particles with increasing distance from the road. It should be noted that magnetic susceptibility enhanced again from 55 to 100 m to form local magnetic highs around the periphery (Figs. 2 and 3). The localized, narrow and linear anomalies represent enhancement of magnetic susceptibility parallel to the street alignment, typical characteristics of traffic-related particles (Hoffmann et al. 1999; Zechmeister et al. 2005; Bućko et al. 2010; Werkenthin et al. 2014; Ma et al. 2016; Nikolaeva et al. 2017; Pierwola et al. 2020).

A few magnetic hotspots around the periphery with intensities of magnetic susceptibility near or over 50 × 10–5 SI might represent the possible input of magnetic material from the nearby alleys if we stick only to the traffic sources (Figs. 2 and 3). However, more realistic explanation would be the difference in topographic elevation and surface morphology. Indeed, local magnetic highs are located in lower elevation, where decline trend of elevation stops (Fig. 3). It is possible that magnetic particles were continuously accumulated in these low elevations over time.

In nature, anthropogenic pollutants occurred in solid, liquid, and gas phases or a combination of multiple phases. Because of on-going water cycling events (i.e., rain or snow), fluidal phases of pollutants were washed away and/or penetrated into the ground. As a result, only the solid phase will dominate the magnetic signal in the topsoil. The solid phase may include pollutants produced mostly by abrasion and corrosion. On the basis of crystal morphology, it is likely that angular iron-rich oxide particles were possibly originated from traffic-related process or natural source (Fig. 7b). It was well documented that vehicles produce angular magnetite particles via non-exhaust emissions as a result of abrasion of brake linings, as well as the abrasion or corrosion of vehicle engine and body work (e.g., Matzka and Maher 1999; Bućko et al. 2010; Bourliva et al. 2016). Particles in rounded shapes were unlikely to represent natural minerals, but were produced by fossil fuel combustion process considering their morphological characteristics (Fig. 7b). Iron-oxide spherules are definitely high-temperature origin, characteristic of vehicle exhaust emission or fly ash particles derived from fossil fuel combustion process (e.g., Spiteri et al. 2005; Goodarzi 2006; Jordanova et al. 2008). These pollutants are probably transported from western coastal industrial areas of South Korea and eastern coastal industrial areas of China (e.g., Kim et al. 2012; Baatar et al. 2017; Ha et al. 2017).

Magnetic minerals are inherited from the parent materials of soils throughout the process of weathering and pedogenesis. Urbanization process accompanies steady accumulation of heavy metals in association with magnetic particles to the soils (Karimi et al. 2011; Dankoub et al. 2012). Influence of pedogenic production of secondary ferrimagnetic material also needs to be counted. Considering meteorological data (e.g., annual rainfalls) and topographic information (e.g., elevation difference), pedogenesis was possibly disturbed continuously under pasture. It should be noted that pedogenic particles are usually much smaller than angular magnetites observed in the present study (Fig. 7b), and usually regular in shape, because they were formed in natural conditions (e.g., Bućko et al. 2010, 2011; Bourliva et al. 2016, 2018). Hence, pedogenic production of secondary ferrimagnetic material was probably insignificant, although we cannot completely rule out the possibility.

Rock magnetic characteristics

Because weak-field properties (e.g., χARM and χfd) are sensitive to fine-grained ferrimagnetic particles, they have been commonly used as granulometric indicators in rock magnetism and environmental magnetism. χfd did not show any significant correlation with IRM-related parameters (Fig. 4), simply because our soil samples were virtually devoid of SP grains (Table 2). Similarly, χARM lacked correlation with IRM-related parameters, also supporting near absence of extremely fine particles (Table 2). Hence, we will focus only on the mutual correlation between high-field IRM-relevant parameters.

S-ratios represent the fraction of lower coercivity minerals to higher coercivity minerals (e.g., Salomé and Meynadier 2004; Liu et al. 2007). Values of S0.1 T, S− 0.1 T, and S− 0.3 T ranged from 0.502 to 0.834, 0.683 to 0.896, and 0.938 to 0.991, respectively (Table 1). In particular, values of S− 0.3 T were extremely high, indicative of a low content of high coercivity minerals. This agrees well with the fact that the IRM acquisitions were saturated by 0.3 T in all samples (Fig. 5).

A total of 25 pairs out of 66 possible combinations yielded significant results with p < 0.01 (Table 2). Among ten rock magnetic factors used, Bcr was chosen as the common parameters in all correlation diagrams (Fig. 8). It is evident that Bcr showed significant correlation with seven other factors (Table 2). For instance, there is a correlation between MDF and Bcr (R2 = 0.5711, Fig. 8a), BR and Bcr (R2 = 0.7711, Fig. 8b), MIRM and Bcr (R2 = 0.7223, Fig. 8c), P and Bcr (R2 = 0.2809, Fig. 8d), and S− 0.1 T and Bcr (R2 = 0.6164, Fig. 8e). As the values of R2 were relatively smaller, it must be regarded that MDF (Fig. 8a) and P (Fig. 8d) as less diagnostic to Bcr than other parameters of BR, MIRM, or S− 0.1 T. No significant correlation between S− 0.3 T and Bcr (R2 = 0.0303, Fig. 8f) was observed. It is apparent that S− 0.3 T is not correlated with all other factors, simply because S− 0.3 T was too high to show any significant trend (Table 2). Insensitivity of S− 0.3 T in this study was somewhat anticipated as high coercivity material is nearly absent in our soil samples (Table 1).

Fig. 8
figure 8

Relation between various magnetic parameters including a median destructive field (MDF) and coercivity of remanence (Bcr), b crossover field (BR) and Bcr, c MIRM and Bcr, d crossover point (P) and Bcr, e S− 0.1 T and Bcr, and f S− 0.3 T and Bcr

Full size image

The classical rock magnetic approach requires modification to incorporate more objective IRM unmixing analysis developed by Kruiver and Passier (2001). We applied the MaxUnMix (Maxbauer et al. 2016, 2017), which is an upgraded semi-automated version of the IRM unmixing analysis (Kruiver and Passier 2001). IRM component analysis yielded one or two magnetic components (Fig. 9). For example, sample GCD31 had 63.0% of softer component (21.3 mT) and 37.0% of harder component (81.3 mT) (Fig. 9a). Results from GCD40 (Fig. 9b) were similar to those of GCD31. Nearly equal contribution from softer component and harder component was observed for GCA05 (Fig. 9c), GCB18 (Fig. 9d), and GCB29 (Fig. 9e). In these samples, dual coercivity peaks were more prominent (Fig. 9c–e) than those for GCD31 or GCD40. Of course, distributions of dual coercivity components were not clearly separable as to blur overlapped coercivity intervals (Fig. 9a–e). Results from GCB11 (Fig. 9f) deserved special attention, because it showed the largest Bcr values (Fig. 8). There was only a single component for GCB11 whose mean coercivity was peaked at 90.7 mT (Fig. 9f).

Fig. 9
figure 9

Coercivity spectra of a GCD31, b GCD40, c GCA05, d GCB18, e GCB29, and f GCB11 determined from MaxUnMix software package (Maxbauer et al. 2016, 2017)

Full size image

What controls the magnetic coercivity spectra in the soil samples? Results from MaxUnMix are well-explained with two distributions of low coercivity phase (21.3–36.4 mT) and high coercivity phase (81.3–151.1 mT). Exceptional results for GCB11 can be regarded as a single coercivity distribution of entirely high-coercivity phase (Fig. 9f). However, a linkage between classical IRM-related parameters (Figs. 5, 6, 7 and 8) and IRM component analysis using MaxUnMix (Fig. 9) was far from being a simple linear relation. The bulk of the observations are consistent with the notion that both classical IRM-related parameters (i.e., S ratios) and IRM component analysis using MaxUnMix are imperfect as a fine probe of granulometry. As the thermal demagnetization of composite IRM data yielded magnetite as the sole remanence carrier (Fig. 7a), it is possible that particle size distribution induced dual coercivity distribution in the soils samples. To this end, it is likely that the softer component was dominated by a lower coercivity component which is interpreted to be angular magnetite originated from exhaust emissions. On the other hand, the harder component was dominated by a high coercivity component that is fine-grained spherical magnetite resulted from fossil fuel combustion process.

Conclusions

Two dimensional map of magnetic susceptibility was constructed using the magnetic susceptibility data measured on 238 topsoil samples from six profiles in Daejeon, South Korea. High values of magnetic susceptibility were narrowly concentrated adjacent to the street. Magnetic susceptibility decreased away from the street. Presence of magnetic hotspots around the periphery with intensities of magnetic susceptibility near or over 50 × 10–5 SI might represent the difference in topographic elevation and surface morphology, because magnetic particles were continuously accumulated in low elevations over time. The maximum unblocking temperatures of composite IRMs were 580 °C, suggesting that the magnetic mineralogy was dominated by magnetite. IRM component analysis showed dual coercivity peaks in most samples. It is likely that the softer component was dominated by coarse-grained angular magnetite originated from vehicle non-exhaust emissions. On the other hand, the harder component was dominated by fine-grained spherical magnetite resulted from exhaust emissions.