Journal of Applied Geophysics 107 (2014) 36–44

Contents lists available at ScienceDirect

Journal of Applied Geophysics

j ourna l homepage: www.e lsev ie r .com/ locate / j appgeo

Magnetic properties of agricultural soil in the Pearl River Delta, SouthChina — Spatial distribution and influencing factor analysis

Yong Bian a,b, Tingping Ouyang a,⁎, Zhaoyu Zhu a, Ningsheng Huang a, Hongfu Wan c, Mingkun Li a,b

a Key Laboratory of Marginal Sea Geology, Guangzhou Institute of Geochemistry, Chinese Academy of Sciences, Guangzhou 510640, Chinab University of Chinese Academy of Sciences, Beijing 100049, Chinac Guangdong Institute of Eco-Environment and Soil Sciences, Guangzhou 510650, China

⁎ Corresponding author. Tel.: +86 20 85290258; fax: +E-mail address: oyangtp@gig.ac.cn (T. Ouyang).

http://dx.doi.org/10.1016/j.jappgeo.2014.05.0030926-9851/© 2014 Elsevier B.V. All rights reserved.

a b s t r a c t

a r t i c l e i n f o

Article history:Received 13 November 2013Accepted 5 May 2014Available online 18 May 2014

Keywords:Magnetic propertiesAgricultural soilSpatial distributionInfluencing factorPearl River Delta

Environmentalmagnetism has beenwidely applied to soil science due to its speediness, non-destructiveness andcost-effectiveness. However, themagnetic investigation of agricultural soil, so closely related to humanactivity, islimited, most probably because of its complexity. Here we present a magnetic investigation of 301 agriculturalsoil samples collected from the Pearl River Delta (PRD, 112°E–115°E and 22°N–24°N), China. The results showedthat both low and high coercivity magnetic minerals coexist in agricultural soil. The values of concentration-dependent parameters, low-field susceptibility (χlf), anhysteretic remanence magnetization susceptibility(χARM), and saturation isothermal remanence magnetization (SIRM) were much higher in the PRD plain thanin the surrounding areas. The S-ratio (S−300) showed a similar spatial pattern to the aforementioned parameters.By contrast, frequency-dependent susceptibility (χfd%) and χARM/SIRMwere higher in the surrounding hilly andmountainous areas than in the PRD plain. Natural and anthropogenic factors such as parent material, soil typeand cultivation methods play important roles in determining agricultural soil magnetic properties. Magneticminerals were coarser grained and overall indicated higher concentrations in soils from river alluvium and de-positedmaterials. Soilswhich had suffered long-termwater submergence have the lowestmagneticmineral con-centration, a result consistent with previous studies. The magnetic properties of agricultural soils are stronglyinfluenced by cultivation methods. Other human activities, such as industrial development and concomitantemitted pollutants, might have had an additional impact on the magnetic properties of agricultural soil.

© 2014 Elsevier B.V. All rights reserved.

1. Introduction

The recently developed environmentalmagnetismmethod is a high-speed, non-destructive and cost-effective technique (Verosub andRoberts, 1995), which has been widely applied in Earth and environ-mental sciences. Many previous studies have discussed sedimentaryprovenance and paleoenvironmental change based on the magnetic in-vestigation of lacustrine and marine sediments and loess/paleosol se-quences (Chen et al., 1999; Hu et al., 2001; Liu et al., 2003; Hounslowand Morton, 2004; Zhang et al., 2012b). In addition, soil magnetism re-search has ventured into heavy metal pollution (Gautam et al., 2004;Gladysheva et al., 2007; Hu et al., 2007) and pedogenesis (Grisonet al., 2011; Lu et al., 2012a).

The magnetic characteristics of soil have attracted a great deal of at-tention in the field of environmental magnetism, because they aregreatly impacted by diverse natural and anthropogenic processes. Forexample, soil developed from basalt usually has much higher suscepti-bility than other bedrock types because of its high background contentof iron oxides (Thompson and Oldfield, 1986; Lu, 2000b; Wang et al.,

86 20 85290278.

2000; Rao et al., 2007). Previous studies have concluded that, overall,parent material is a critical factor determining the type and concentra-tion of magnetic minerals (Fialova et al., 2006; Blundell et al., 2009a).Other natural factors, such as climate, topography, water regime, organ-isms, etc. can lead to variousmagnetic characteristics in soils (Lu, 2000a;Hanesch and Scholger, 2005; Lu et al., 2012a). Besides natural factors,human activitymay affect soilmagnetism inmanyways. Anthropogenicpollutants containing magnetic fractions, mainly derived from industryand transportation vehicles, can enhance soil magnetism in the surfacelayer (Blundell et al., 2009b; Bucko et al., 2010; Karimi et al., 2011;Meena et al., 2011). Cultivation may be another factor influencing soilmagnetic characteristics. Susceptibilities are usually lower in yearlyoverturned soils than in uncultivated ones (Durza, 1999; Magiera andZawadzki, 2007).

Agricultural soil is a soil category closely associated with humanactivities. However, magnetic research focusing on agricultural soil islimited. Magiera and Zawadzki (2007) revealed thatmagnetic suscepti-bility in arable land is statistically significantly lower than that in forestland, due to the dilution of magnetic minerals with non-magnetic com-ponents, such as organic matter. Magnetic methods can be used as aproxy to detect heavy metal contamination in agricultural soil pollutedby either wastewater or atmospheric pollutants (Duan et al., 2010;

37Y. Bian et al. / Journal of Applied Geophysics 107 (2014) 36–44

Zhang et al., 2013). Duan et al. (2010) also suggested that the use of pes-ticides and fertilizers had no effect on the magnetic properties found inpolluted topsoil. In China, paddy soil is the agricultural soil which at-tractsmost environmentalmagnetism studies. Yan et al. (2011) pointedout that paddy soil is an ideal soil type for heavy metal detection usingmagnetic proxies, due to its low magnetic background and high homo-geneity in the top layer. Lu et al. (2012b) and Han and Zhang (2013)conducted detailed investigations into the magnetic properties ofpaddy soil using both magnetic and non-magnetic methods, in orderto advance knowledge of the relation between its pedogenesis andsoil magnetism.

The magnetic characteristics of agricultural surface soils of the PearlRiver Delta (PRD) are the subject of this study. The PRD, with longitudeand latitude ranging from 112°E to 115°E and 22°N to 24°N, respective-ly, is located in the subtropical Guangdong Province, South China. Fig. 1displays the location of the PRDand the elevation in this region. The PRDplain area denotes the areas in the core location of the region with ele-vations lower than 20 m and with major rivers flowing through them.The Pearl River consists of three main branches: the West River, theNorth River and the East River. After joining together in the PRD drain-age basin, they flow into the South China Sea through eight estuaries(Weng, 2007, Fig. 1). Hills andmountains arewidely distributed aroundthePearl River alluvial plain,whichhas been in formation since themid-late Quaternary. Since the ‘Informing and Opening Policy’ launched inthe late 1970s, the PRD has progressively become one of the most im-portant economic regions in China, with manufacturing industries de-veloping rapidly. As a result, the population of the studied area hasincreased substantially. According to the results of the Sixth NationalPopulation Census of the country (Statistics Bureau of GuangdongProvince, Office for the Population Census of Guangdong Province,2012), there were nearly 54 million residents in this 41,500 km2 regionin 2010. Moreover, this region is quite suitable for agriculture due to itswarm and wet climate. Agriculture has been developed for over2000 years in the PRD (Weng, 2007). In plain areas, rice and abundant

Fig. 1. Geographical location of the Pearl River De

species of vegetables are cropped and fishponds are widely distributed.On the other hand, hilly areas are mainly covered by fruit orchards orforest. However, with the growth of both the economy and population,a lot of agricultural land has been replaced by manufacturing and resi-dential housing, associatedwith increasing pollutant emissions. In sum-mary, it can be supposed that the magnetic properties of agriculturalsoils in the PRD are influenced by various factors. The main aim of thepresent study is to identify the principal factors influencing agriculturalsoil magnetism in the PRD. The results presented here provide somepointers on which to build further and deeper research in this field.

2. Data and methods

2.1. Sample acquisition

The 301 studied agricultural soil samples were collected from thePRD by the Guangdong Institute of Eco-Environment and Soil Sciencesduring 2006 and 2009. The sampling siteswere relatively evenly distrib-uted across the PRD (Fig. 1). Samples were obtained on well-developedagricultural soils within a depth of 20 cm. All samples were air-dried atroom temperature and ground into a powder. The material was thenweighed and packed into non-magnetic plastic cubes (with a volumeof 8 cm3). In addition to latitude and longitude, information on parentmaterial, soil type, topography, tillage type, and irrigation type was de-scribed for every soil sample. All these items of information are classi-fied in Table 1 and allow for a grouping of variables for statisticalanalyses.

2.2. Magnetic parameters and measurements

Low-fieldmagnetic susceptibility, which is the composite reflection ofall magneticminerals within a specimen, is themost extensively used pa-rameter in environmental magnetism (Thompson and Oldfield, 1986).Magnetic susceptibility was measured under low fields (200 m/A) at

lta (PRD) and distribution of sampling sites.

Table 1Description and classification of soil samples.

Factor Category Number of samples Essential description

Parent material Eluvial and deluvial materials on igneous rocks 98 Dominantly granite in the study area; ‘igneous rocks’ for short.Eluvial and deluvial materials on sedimentary rocks 62 Dominantly arenaceous shale in the study area; ‘sedimentary rocks’ for short.River alluvial and deposited materials 134 The latest alluvium formed during Quaternary; ‘river alluviums’ for short.Marine sedimentary material 6Other parent material 1

Soil type Paddy soil 100 An artificial soil type formed after long-time cultivation of natural soil inwaterlogged conditions.

Alluvial soil 116 A soil type developed from river alluvium, deltaic deposit or valley alluvium,also cultivated intensively.

Lateritic red soil 81 A soil type formed after intense allitic process, mainly in hilly area covered by forest.Calcareous soil 2Seashore saline soil 2

Topography Plain 216 Elevation lower than 100 m.Hill 51 Elevation between 100 m and 500 m.Mountain 34 Elevation higher than 500 m.

Tillage type Flatbusting and plowing tillage 57 Agricultural soil experiencing frequently overturning.Ridge tillage 150 Ridged soil with crop growing on it; ridges alternating with channels.No tillage 94 Soil growing crops or plants without tillage.

Irrigation type Flood irrigation 57 Soil irrigated by water overflowed on flat fields.Other irrigation type 186 Water-saving types, e.g. sprinkling irrigation and micro irrigation, etc.No irrigation 58 Fields without irrigation facility.

Note: Flatbusting and plowing tillage and flood irrigation information refers to the same samples, since these two types of cultivation are simultaneously adopted in the PRD. Themarinesedimentary material, other parent material, calcareous soil and seashore saline soil categories include only a few samples, and thus cannot be considered representative of thesecategories for thewhole region. Therefore, samples falling into these categories were not used in category-dependent statistical procedures, including non-parametric testing, correlationanalysis and multiple correspondence analysis.

38 Y. Bian et al. / Journal of Applied Geophysics 107 (2014) 36–44

both low (976 Hz, κlf) and high frequency (15,616 Hz, κhf) using aKappabridge MFK1-FA (AGICO). Mass-specific susceptibility (χ) was cal-culated by dividing κ through the density of the samples. Frequency-dependent susceptibility [χfd% = 100 × (κlf − κhf) / κlf] is a widely usednormalized parameter reflecting magnetic particles with grain sizesclose to the superparamagnetic/single domain (SP/SD) boundary(Verosub and Roberts, 1995).

Anhysteretic remanence magnetization (ARM) was imparted toeach sample at an AF peak field of 100 mT with a 0.05 mT DC biasfield using a D-2000 A.F. Demagnetizer and then measured on a JR-6ASpinner Magnetometer. ARM was then normalized by the bias fieldand converted to mass-specific susceptibility of ARM (χARM). This pa-rameter is also related to magnetic mineral abundancewithin a sample.It is also very sensitive to the concentration of SD particles (Verosub andRoberts, 1995; Evans and Heller, 2003).

Specimens underwent isothermal remanent magnetization (IRM)by applying a 2 T field pulse with an IM-10-30 Impulse Magnetizer,and were then measured with a JR-6A Spinner Magnetometer. ThisIRM is regarded as saturation IRM (SIRM) in the present study. SIRMmainly reflects the amount of magnetic minerals within a specimen(Thompson and Oldfield, 1986). SIRM was also transformed to mass-specific form. After acquiring SIRM values, specimens were subjectedto a back field of 300 mT and the resulting remanence was measuredand denoted as IRM−300 mT. The S-ratio (S−300), which reflects the rel-ative concentration of high-coercivity antiferromagnetic minerals with-in a sample (King and Channell, 1991), was then calculated using theformula S−300 = − IRM−300 mT / SIRM. S-ratio represents the relativeconcentrations of ferrimagnetic/antimagneticmineralswithin a sample.It indicates that magnetic minerals in a sample are all low-coercivitywhen it is equal to 1. S-ratio decreases with the increasing abundanceof high-coercivity magnetic minerals. In addition, the χARM/SIRM,which reflects the magnetic grain size within a sample (Evans andHeller, 2003), was calculated. Larger values of this ratio indicate a small-er grain size distribution.

35 representative samples were selected for magnetic mineralo-gy analysis. IRM acquisition curves were measured from 1 mT to2000mT using IM-10-30 and JR-6A. Temperature-dependent magneticsusceptibilities (κ–T curves) of these samples were measured using a

CS4 high temperature unit attached to the Kappabridge MFK1-FA(AGICO) in an argon atmosphere. Each specimen was heated from am-bient temperature to 700 °C and cooled back to 40 °C.

2.3. Statistical analysis

Magnetic parameters were interpolated using ordinary kriging inArcGIS software, in order to reveal the spatial trends of variables withinthe study area. Spearman's rho method was used to compute bivariatecorrelation coefficients among the variables since the results of the var-iables do not achieve a normal distribution. The non-parametric test forindependent samples was used for significance testing. Multiple corre-spondence analysis was adopted to investigate further factors influenc-ing magnetic properties. All statistical procedures were carried outusing SPSS 16.0 software.

3. Results

3.1. Magnetic mineralogical properties

Among the chosen samples for magnetic mineralogical measure-ment, three representative samples were selected, and their κ–Tcurves are shown in Fig. 2a. Samples I and II's soil types are alluvialsoil and lateritic red soil, respectively. The κ–T curves for these twosamples are typical for most samples taken from all soil categories.Susceptibility at room temperature is determined by ferrimagneticmagnetite and/or maghemite (Thompson and Oldfield, 1986). Theheating curves of samples I and II display an obvious increase ataround 250 °C and peak at 280 °C. This increase might be attributedto the existence of magnetic iron hydroxides, which convert toferrimagnetic maghemite when heated to over 250 °C (Oches andBanerjee, 1996). Having peaked at 280 °C, the curves decrease untilthe samples are heated to 400 °C; this may be the result of an inver-sion of metastable maghemite to stable antiferromagnetic hematite(Oches and Banerjee, 1996). There is a second peak at about 500 °Cand then susceptibility drops rapidly to a small value close to zerountil about 580 °C indicating the presence of magnetite (Dunlopand Özdemir, 1997). This second peak of susceptibility might be

Fig. 2. Magnetic mineralogical measurement results for representative samples. a) κ–T curves; b) IRM acquisition curves.

39Y. Bian et al. / Journal of Applied Geophysics 107 (2014) 36–44

raised by the conversion of a hematite and clay minerals admixtureor by the conversion of iron containing silicates to magnetite in anargon atmosphere (Oches and Banerjee, 1996; Zhang et al., 2012a).The κ–T curve pattern for sample III is found in only a few samples,and all these samples are classified as alluvial soil. There is only onepeak at ca. 500 °C on the heating curve of sample III, and the Curiepoint of 580 °C also confirms the presence of magnetite. The highersusceptibility during the cooling process confirmed the generationof magnetite phases during the heating procedure.

Fig. 2b shows the IRM acquisition curves of the three samples. It wasreadily established that samples I and II have different IRM acquisitioncurve patterns. For Sample I, the IRM/SIRM values at 100 mT and300 mT are 0.8 and 0.9, respectively. However, for Sample II, the IRMvalue at 100 mT is only about a half of the SIRM value, and IRM/SIRMvalues at 300mTare only 0.6. This implies that the relative concentrationsof the two magnetic components (low-coercivity and high-coercivity)within these two samples are very different although they exhibit similarhigh temperature behavior. Sample III also contains a considerable con-tent of high coercivity magnetic phases. In addition, the IRMs of thesethree samples are not saturated even at 2 T, indicating that high coercivitymagnetic minerals are present in these samples. The coexistence of highand low coercivity magnetic phases and their differing constitutionsmay suggest a wide range of magnetic parameters within the regionstudied.

3.2. Spatial distribution of magnetic parameters

Fig. 3 displays the spatial distribution of the investigated magneticparameters. It can readily be seen from the maps that χlf, χARM, andSIRM display very similar spatial distribution characteristics (Fig. 3ato c). Specimens with higher values of these three parameters weremainly sampled from the PRD plain area, such as Shunde, Zhuhai, andPanyu, covering the eight estuaries of the Pearl River drainagebasin. The distribution characteristics of these three parameters in-dicate that samples collected from the PRD plain area contain higherconcentrations of magnetic minerals. On the other hand, χfd% andχARM/SIRM show another spatial pattern in that high-value areasare located in the hilly regions around the PRD plain, while theplain itself is a low-value area (Fig. 3d and e). This implies that mag-netic particles within samples collected from the hilly regions arefiner. Specimens from the PRD plain area have relatively high S−300

values (greater than 0.8, Fig. 3f), indicating that low-coercivity min-erals dominate in these samples.

4. Discussion

4.1. Magnetic properties of different soil categories

Results of non-parametric tests for independent samples are illus-trated here as error bar charts that display means and the 95% confi-dence intervals of the variables (Fig. 4).

χlf and χARM convey no significant difference among the three parentmaterial categories (Fig. 4). However, a slightly increasing trend fromigneous and sedimentary rocks to river alluviums is identified in thechart. This trend implies the presence of some influencing factors onχlf and χARM, such as the lithogenicmineral concentrations found in var-ious bedrock or sediment types. These differences have also been foundby other researchers (Hanesch and Scholger, 2005; Fialova et al., 2006).In contrast, SIRM of soils derived from river alluvium was significantlyhigher than the other two categories, indicating that themagnetic min-eral concentration of remanence-bearing minerals is higher in alluvialparent material. On the other hand, the two grain size related parame-ters, χfd% andχARM/SIRM, display significant differences amongdifferentparent materials. Values of these two parameters of soils derived fromriver alluviums were lower than those of soils derived from igneousand sedimentary rocks, indicating that magnetic particles within soilsderived from river alluviums contain fewer SP and SDparticles.Magnet-ic grain size differences between igneous and sedimentary rocks are at-tributed to the various inherent properties of rock and other materials.As sedimentary rocks in the study area are mostly shale, which is con-solidated mainly from clay, sedimentary rocks may contain more finemagnetic particles than igneous rocks (mainly granite), in which thesize of the mineral grains is usually greater. S−300 values for the threekinds of parent material do not show variation significantly, implyingthat differences in parent material do not account for the relativeabundance of high coercivity magnetic minerals found in extremely-weathered surface soils within the study area.

When comparisons are applied to soil types, alluvial soils, which areclosely related to river alluvium, shows magnetic properties that arevery similar to those in river alluviums (Fig. 4). The comparison be-tween paddy soil and lateritic red soil can be regarded as an indicatorof the contrast between waterlogged and well-drained conditions. Theχlf and χARM values of lateritic red soil are significantly higher thanthose of paddy soil, which can be attributed to magnetic mineral lossraised by waterlogging (Lu, 2000a; Hanesch and Scholger, 2005). Atthe same time, the lateritic red soil category has higher χfd% and χARM/SIRM values than paddy soil, implying that magnetic particles withinwell-drained soils are finer than those in waterlogged soils. This could

Fig. 3. Spatial distribution ofmagnetic parameters generatedby kriging interpolation, a)χlf (10−8m3/kg), b)χARM (10−8m3/kg), c) SIRM(10−3 Am2/kg), d)χfd%, e)χARM/SIRM(10−3 m/A), andf) S−300.

40 Y. Bian et al. / Journal of Applied Geophysics 107 (2014) 36–44

Fig. 4. Error bar charts of magnetic parameters, showing the different categories of factors and indicating the 95% confidence intervals of variables. The marker * indicates significant dif-ferences among the defined categories analyzed by non-parametric testing.

41Y. Bian et al. / Journal of Applied Geophysics 107 (2014) 36–44

be due to the fact that finer magnetic particles are more easily lost inwaterlogged conditions (Lu et al., 2012b). Correlation coefficients be-tween χlf and χfd% are 0.121, 0.212, and 0.499 for alluvial soil, paddysoil and lateritic red soil, respectively. The differences between these co-efficients imply that themagnetic susceptibility of the well-drained andrelatively natural lateritic red soil type correlatesmore closelywithfinergrained particles than with paddy soil and alluvial soil.

When comparing topographic types (Fig. 4), the Plain category,which accounts for most of the samples, has the lowest χfd% and χARM/SIRM, while magnetic concentration related parameters do not showsignificant differences among topographic types. This implies that, inthe whole PRD region, topography does not play an evident role in af-fecting magnetic concentration, but relates significantly to particlesize. In fact, in different terrains, cultivation activities in differing catego-ries of agricultural land can vary substantially, which may explain themagnetic particle size variation among topographic types. Thus, the fac-tor influencingmagnetic grain sizemay be the intensity of human activ-ities rather than topography itself.

Anthropogenic influencing factors, including both tillage and irriga-tion types, cause significant differences in magnetic properties (Fig. 4).Flatbusting and plowing tillage, as well as flood irrigation, represent in-tense cultivation and waterlogged conditions that are similar to paddy

soil, so these soils, like paddy soil, show the lowest values of χlf, χARM,and SIRM. χfd% and χARM/SIRMvalues for flatbusting and plowing tillageand ridge tillage, aswell as forflood irrigation and other irrigation types,are evidently lower than for areas with no tillage or irrigation, whichmeans thatmagnetic particles within the former categories are relative-ly coarser. This may be the result of differences in soil formation causedby cultivation, or it may indicate that the smaller magnetic particles ap-pear to dissolve more readily in conditions of such intensive cultivation.χlf and χARM values of no tillage and no irrigation soils that are character-ized by less intensive farming, are similar to soils that were subjected toridge tillage and other irrigation types. However, SIRM varies signifi-cantly among the three sub-groups. The SIRM of other irrigation typesand ridge tillage is the highest, followed by no tillage and no irrigationtypes; flatbusting and plowing tillage, and flood irrigation, evince thelowest SIRM. The differences in performance between SIRM and theother twomagnetic concentration-related parameters, χlf and χARM, indi-cate the complexities in the variation ofmagnetic properties in agricultur-al soil. When comparing S−300 values between categories of tillage typeand irrigation type, it was found that ridge tillage and other irrigationtypes were highest in S-ratio, which suggests that cultivation methodsalso have an evident effect on variations in the relative abundance of fer-rimagnetic minerals in the soil.

Table 2Bivariate non-parametric Spearman's correlation coefficients.

Variable χlf χARM SIRM χfd% χARM/SIRM S−300

χlf 1χARM 0.933a 1SIRM 0.896a 0.882a 1χfd% 0.277a 0.352a 0.086 1χARM/SIRM −0.013 0.155a −0.281a 0.601a 1S−300 0.517a 0.500a 0.453a 0.124b 0.025 1

a Means correlations are significant at the 0.01 level (2-tailed).b Means correlations are significant at the 0.05 level (2-tailed).

Fig. 5.MCA results with categories projected on factorial plane D1/D2. These two factorsaccount for 85% of the total variance. Dimension 1 (Factor 1) could be a proxy of cultivationintensity; cultivation intensity increases gradually from the left to the right. Dimension 2(Factor 2) could approximately reflect the concentration ofmagnetic minerals. Categoriesthat are distributed next to each other are correlated closely. A category that is distributeddistantly from the origin possessesmore obviousmagnetic characteristics and its magnet-ic parameters are thus more significantly influenced by factors reflected by Dimension 1(cultivation intensity) and Dimension 2 (magnetic mineral concentration).

42 Y. Bian et al. / Journal of Applied Geophysics 107 (2014) 36–44

4.2. Correlation between magnetic parameters

Table 2 lists the bivariate correlation coefficients between the mag-netic parameters for all samples. χlf, χARM, and SIRM are shown to bevery closely correlated, due to similar reflections of magnetic mineralabundance. χARM/SIRM shows a high correlation coefficient with χfd%,suggesting that fine grainmagnetic particles have the same origin in ag-ricultural soil. Moreover, S−300 values positively correlate with χlf, χARM,and SIRM. The correlation coefficient between χARM/SIRM and χlf is veryclose to zero among all samples, indicating that magnetic mineral con-centrations and grain size are relatively independently impacted by dif-ferent factors.

Besides establishing the correlation coefficients for all samples, cor-relation analysis was also carried out for each category of soil (resultsare not shown). Correlations between χlf, χARM, and SIRM are significantfor all categories, similar to that of the whole sample set. In addition,χfd% and χARM/SIRM are also significantly correlated for all categories.However, the significance of other correlation coefficients varies be-tween categories. Although χARM/SIRM and χlf are not significantly cor-related for the whole set of samples, they are positively correlated forthe soils in areas which experience less intense cultivation, such as re-gions dominated by sedimentary rocks, hills, lateritic red soil and no till-age. These two parameters are even negatively correlated for plain andflood irrigation soils. Different from the correlation of the whole sampleset, S−300 does not show significant correlationwith χlf, χARM, and SIRMfor sedimentary rocks, hills, mountains, lateritic red soil, and areas of noirrigation. Moreover, there is no significant correlation between χfd%and χlf, χARM for river alluviums, alluvial soil and ridge tillage. Therefore,it may be inferred that soil category differences can have a significantinfluence on magnetic phase constitution and grain size, leading to arange of correlations between magnetic parameters for different soilcategories.

4.3. Analysis of influencing factors using the multiple correspondenceanalysis method

Multiple correspondence analysis (MCA) was applied to gain a fur-ther understanding of the factors influencing the magnetic propertiesof agricultural soils. This method is suitable for samples with multiplenominal variables and effective for investigating the relations betweendifferent categories. It has been successfully used in geoscience andsoil science (Reis et al., 2007; Feuillet, 2011). Before using this method,a categorization of numeric variables was carried out according to thefollowing rules.

As the variables χlf, χARM, and SIRM follow a log-normal distribution,they were transformed to logarithmic values for the even sample num-bers of new categories. Aswell as the other three variables, theywere allgiven new nominal values for each sample. Samples that are less thanthe mean value and with differences greater than one deviation weredesignated as ‘1’; those less than the mean value and with differencesless than one deviation were designated as ‘2’. Similarly, samples great-er than the mean value and with differences less and greater than onedeviation were designated as ‘3’ and ‘4’, respectively. Therefore, thenew values from 1 to 4 mean that the original values for each magnetic

parameter vary from small to large. Together with the five soil categoryvariables, a total of 11 category variables were considered using MCA.The MCA results are shown in Fig. 5. It can be easily noted that soil cat-egories such asno irrigation, no tillage, andmountain, distributed on theleft of the figure, represent the types with weaker cultivation activities,and those on the right, the contrary. So Dimension 1 could be a proxy ofcultivation intensity. On the other hand, categories labeled ‘1’ for mag-netic mineral concentration related variables (with solid markers) arenear the bottom of the figure, and those labeled ‘4’ are near the top. SoDimension 2 could approximately represent magnetic mineralabundance.

Soil categories on the left are not far from the horizontal axis, fromwhich itmay be inferred that there is amoderate abundance ofmagnet-ic minerals in this region's relatively natural soils. On the contrary, soilcategories experiencing relatively intense cultivation are distributedmore widely along the vertical axis, which implies that human cultiva-tion is mainly responsible for magnetic mineral abundance variationswithin the agricultural soil. The categories of lowest magnetic mineralconcentration (solid markers labeled ‘1’) are very close to those ofpaddy soil, flood irrigation, flatbusting and plow tillage, implying thatthese soil categories contain the least magnetic minerals. Furthermore,these categories containing the least magnetic minerals are located far-thest from the vertical axis, suggesting that they are influenced by themost intense cultivation. However, the solid and cross markers labeled‘4’ are not proximate to any soil categories, indicating that sampleswith the most absolute and relative ferrimagnetic minerals are not in-fluenced by any specific soil category factor. These magnetic mineral-abundant samples might indeed be influenced by other sources, e.g.industry. Thus, anthropogenic emissions should be taken into accountin any further analysis.

The categories with the finest magnetic particles (hollow markerslabeled ‘4’) are clearly located very near to weak-cultivation soil

Fig. 6. Industrial density from 2001 to 2010, expressed as millions of Yuan per km2 for each prefecture-level city and county.

43Y. Bian et al. / Journal of Applied Geophysics 107 (2014) 36–44

categories such as no irrigation, no tillage, mountain, and lateritic redsoil. This implies that such soil categories relate closely to soils withmagnetically finer grain characteristics, coinciding with the resultsshown in the error bar charts in Fig. 4. However, some informationcan be obtained using MCA which cannot be detected using error bars.The plain and hollow circle markers ‘1’ and ‘2’ are very close to eachother, indicating that magnetic particles with SP/SD boundary grainsize are not abundant in the PRD plain, where the majority of samples,and much of the farmland, are located. However, the hollow squaremarker ‘1’, which is representative of the fewest fine magnetic grains,is not proximate to the hollow circle markers ‘1’ and ‘2’. Both these pa-rameters can be indicators ofmagnetic grain size, but theymight also beinfluenced by different factors. This least amount of finemagnetic parti-cles might not be the result of cultivation activity in the PRD plain, butrather may have industrial sources, since industrially-derived magneticparticles are evidently larger than natural ones (Blaha et al., 2011;Meena et al., 2011).

4.4. Relation betweenmagnetic properties andmetal elements in agriculturalsoil

According to the geochemical maps of elements in surface soil(China Geological Survey andMinistry of Land Resources of the People'sRepublic of China, 2011), the elementsMn, Cu, Cr, Co, Cd, Ni, Pb, Zn, etc.,evince high concentrations in the PRD plain, which also happens to be ahigh value area for the χlf, χARM, SIRM and S−300 parameters. It couldtherefore be inferred that magnetic properties may be correlated withthese elements. As indicated by Luo et al. (2012), magnetic susceptibil-ity is correlated with free iron concentration (pFe) in soil, which de-creases as cultivation intensity increases. Data provided by Luo et al.(2012) also shows that free iron concentrations in the top layer ofcultivated lateritic red soil are significantly lower than that in naturallateritic red soil. At the same time, Mo and Mn concentrations showsimilar variations between these soils. That is to say, concentrations ofcertain elements and magnetic minerals might be similarly influencedby cultivation.

Many previous studies have suggested that pollutant inputs, espe-cially heavy metal pollution, usually accompany enhanced magnetismin the topsoil in urban and industrial areas (Gautam et al., 2004;

Wang and Qin, 2005; Blaha et al., 2008; Matysek et al., 2008). In orderto study the relation between soil magnetic properties and anthropo-genic emission, the industrial density for each administrative district,whichwas calculated as gross industrial output value (in unit of millionRMB Yuan, from 2001 to 2010) normalized by corresponding area(Statistics Bureau of Guangdong Province, Guangdong Survey Office ofNational Bureau of Statistics, 2002–2011), is illustrated in Fig. 6. It isclear that the spatial distribution of industrial density is similar to thatof χlf, χARM, SIRM, and S−300 values. Districts with higher industrial out-put are most commonly located in the PRD plain area, where the mag-netic parameters are also higher in value than in the surroundingareas. This similarity in spatial distribution between industrial densityand magnetic parameters may be attributed to the industrially-emittedpollution correlated with magnetic mineral concentrations. It can alsobe seen that estuary zones possess higher magnetic parameter values al-though their industrial densities are not especially high; this may be theresult of magnetic substances and pollutants being transferred in. Pollut-ants can be transferred both by air and by water, and then deposited onthe soil (Wong et al., 2003; Schmidt et al., 2005; Wang et al., 2012).

5. Conclusion

The spatial distribution, and factors influencing, the magnetic prop-erties in agricultural soils collected from the PRDwere investigated. Thefollowing conclusions can be inferred from the results:

1) All agricultural soil samples contain low-coercivity and high-coercivity magneticmineral phases. Concentration-dependentmag-netic parameters, χlf, χARM, and SIRM, show relatively high values inthe plain area of the PRD. Some districts, such as Shunde, Shenzhenand Zhuhai, carry higher values of these parameters. In contrast, thegrain size dependentmagnetic parameters, χfd% and χARM/SIRM, dis-play low values in the plain area and relatively high values in thesurrounding hilly and mountainous areas. The S−300 evinces a simi-lar spatial distribution to those concentration-dependent magneticparameters.

2) Both natural and anthropogenic factors such as parent material, soiltype and cultivation method can affect the magnetic properties ofagricultural soils by way of both magnetic mineral concentrationand grain size. Magnetic minerals within soils derived from river

44 Y. Bian et al. / Journal of Applied Geophysics 107 (2014) 36–44

alluviums were found to be more abundant and contain less finemagnetic grains than the others. Waterlogging and well-drainedconditions associated with soil type and irrigation greatly affectedthe soil's magnetic properties. Soils with long-term water coveragehad the lowest magnetic mineral concentrations. Cultivation maybe another important factor affectingmagnetism. Soils experiencingthe most intense cultivation contain the least magnetic concentra-tions and fine magnetic grains.

3) Human activities, such as industrial development and concomitantemitted pollutants, may be responsible, to some extent, for varia-tions in themagnetism of agricultural soils, as emitted heavy metalsare likely to coexist with anthropogenic magnetic particles. Anthro-pogenic magnetic particles may enhance magnetism and lead to acoarser grain size distribution in the topsoil.

Acknowledgements

Our work has benefitted from the support of the NSFC–GuangdongJoint Fund (Grant No. U1201131), the Science and Technology Project ofGuangdong Province (Grant No. 2011B030500031), and the GuangdongProvince Chinese Academy of Sciences Comprehensive Strategic Cooper-ation Project (Grant Nos. 2012B090400045 & 2011B090300052).

References

Blaha, U., Appel, E., Stanjek, H., 2008. Determination of anthropogenic boundary depth inindustrially polluted soil and semi-quantification of heavymetal loads using magnet-ic susceptibility. Environ. Pollut. 156, 278–289.

Blaha, U., Basavaiah, N., Deenadayalan, K., Borole, D.V., Mohite, R.D., 2011. Onset of industrialpollution recorded in Mumbai mudflat sediments, using integrated magnetic, chemical,Pb-210 dating, and microscopic methods. Environ. Sci. Technol. 45, 686–692.

Blundell, A., Dearing, J.A., Boyle, J.F., Hannam, J.A., 2009a. Controlling factors for the spatialvariability of soil magnetic susceptibility across England and Wales. Earth-Sci. Rev.95, 158–188.

Blundell, A., Hannam, J.A., Dearing, J.A., Boyle, J.F., 2009b. Detecting atmospheric pollutionin surface soils using magnetic measurements: a reappraisal using an England andWales database. Environ. Pollut. 157, 2878–2890.

Bucko, M.S., Magiera, T., Pesonen, L.J., Janus, B., 2010. Magnetic, geochemical, and micro-structural characteristics of road dust on roadsides with different traffic volumes—case study from Finland. Water Air Soil Pollut. 209, 295–306.

Chen, F.H., Shi, Q., Wang, J.M., 1999. Environmental changes documented by sedimenta-tion of Lake Yiema in arid China since the late glaciation. J. Paleolimnol. 22, 159–169.

China Geological Survey, Ministry of Land Resources of the People's Republic of China,2011. Multi-Purpose Regional Geochemistry Atlas: Pearl River Delta Economic Zoneof Guangdong Province, P.R. China. Geological Publishing House, Beijing.

Duan, X.M., Hu, S.Y., Yan, H.T., Blaha, U., Roesler, W., Appel, E., Sun, W.H., 2010. Relation-ship betweenmagnetic parameters and heavy element contents of arable soil arounda steel company, Nanjing. Sci. China Earth Sci. 53, 411–418.

Dunlop, D.J., Özdemir, Ö., 1997. Rock Magnetism: Fundamentals and frontiers. CambridgeUniversity Press, Cambridge.

Durza, O., 1999. Heavy metals contamination and magnetic susceptibility in soils aroundmetallurgical plant. Phys. Chem. Earth A 24, 541–543.

Evans, M.E., Heller, F., 2003. Environmental Magnetism: Principles and Applications ofEnviromagnetics. Academic Press, Oxford.

Feuillet, T., 2011. Statistical analyses of active patterned ground occurrence in the TaillonMassif (Pyrenees, France/Spain). Permafr. Periglac. 22, 228–238.

Fialova, H., Maier, G., Petrovsky, E., Kapicka, A., Boyko, T., Scholger, R., 2006. Magneticproperties of soils from sites with different geological and environmental settings.J. Appl. Geophys. 59, 273–283.

Gautam, P., Blaha, U., Appel, E., Neupane, G., 2004. Environmental magnetic approach to-wards the quantification of pollution in Kathmandu urban area, Nepal. Phys. Chem.Earth 29, 973–984.

Gladysheva, M.A., Ivanov, A.V., Stroganova, M.N., 2007. Detection of technogenically contam-inated soil areas based on their magnetic susceptibility. Eurasian Soil Sci. 40, 215–222.

Grison, H., Petrovsky, E., Jordanova, N., Kapicka, A., 2011. Strongly magnetic soil devel-oped on a non-magnetic rock basement: a case study from NW Bulgaria. Stud.Geophys. Geod. 55, 697–716.

Han, G.Z., Zhang, G.L., 2013. Changes in magnetic properties and their pedogenetic impli-cations for paddy soil chronosequences from different parent materials in southChina. Eur. J. Soil Sci. 64, 435–444.

Hanesch, M., Scholger, R., 2005. The influence of soil type on the magnetic susceptibilitymeasured throughout soil profiles. Geophys. J. Int. 161, 50–56.

Hounslow, M.W., Morton, A.C., 2004. Evaluation of sediment provenance using magneticmineral inclusions in clastic silicates: comparison with heavy mineral analysis. Sedi-ment. Geol. 171, 13–36.

Hu, Y., Oldfield, F., Manalt, F., Beck, C., 2001. The environmental significance of magneticmeasurements of Late Pleistocene and Holocene sediments from the Grand Lacd'Annecy, eastern France. J. Paleolimnol. 25, 193–203.

Hu, X.F., Su, Y., Ye, R., Li, X.Q., Zhang, G.L., 2007. Magnetic properties of the urban soils inShanghai and their environmental implications. Catena 70, 428–436.

Karimi, R., Ayoubi, S., Jalalian, A., Sheikh-Hosseini, A.R., Afyuni, M., 2011. Relationships be-tweenmagnetic susceptibility and heavymetals in urban topsoils in the arid region ofIsfahan, central Iran. J. Appl. Geophys. 74, 1–7.

King, J.W., Channell, J.E.T., 1991. Sedimentary magnetism, environmental magnetism, andmagnetostratigraphy. Rev. Geophys. 29, 358–370.

Liu, H., Zhu, R.X., Li, G.X., 2003. Rock magnetic properties of the fine-grained sediment onthe outer shelf of the East China Sea: implication for provenance. Mar. Geol. 193,195–206.

Lu, S., 2000a. Characterization of subtropical soils by mineral magnetic measurements.Commun. Soil Sci. Plant 31, 1–11.

Lu, S.G., 2000b. Lithological factors affecting magnetic susceptibility of subtropical soils,Zhejiang Province, China. Catena 40, 359–373.

Lu, S.G., Chen, D.J., Wang, S.Y., Liu, Y.D., 2012a. Rock magnetism investigation of highlymagnetic soil developed on calcareous rock in Yun-Gui Plateau, China: evidence forpedogenic magnetic minerals. J. Appl. Geophys. 77, 39–50.

Lu, S.G., Zhu, L., Yu, J.Y., 2012b. Mineral magnetic properties of Chinese paddy soils and itspedogenic implications. Catena 93, 9–17.

Luo, Y., Liu, W., Wu, L., Song, J., Tu, C., Zhang, H., Teng, Y., 2012. The soil of Yangtze RiverDelta and Pearl River Delta Regions and its Environment. Science Press, Beijing (inChinese).

Magiera, T., Zawadzki, J., 2007. Using of high-resolution topsoil magnetic screening for as-sessment of dust deposition: comparison of forest and arable soil datasets. Environ.Monit. Assess. 125, 19–28.

Matysek, D., Raclavska, H., Raclavsky, K., 2008. Correlation between magnetic sus-ceptibility and heavy metal concentrations in forest soils of the eastern CzechRepublic. J. Environ. Eng. Geophys. 13, 13–26.

Meena, N.K., Maiti, S., Shrivastava, A., 2011. Discrimination between anthropogenic(pollution) and lithogenic magnetic fraction in urban soils (Delhi, India) usingenvironmental magnetism. J. Appl. Geophys. 73, 121–129.

Oches, E.A., Banerjee, S.K., 1996. Rock-magnetic proxies of climate change from loess–paleosol sediments of the Czech Republic. Stud. Geophys. Geod. 40, 287–300.

Rao, Z.G., Zhu, Z.Y., Wu, Y., Yang, T., Fu, S.Q., 2007. Character and significance of magneticsusceptibility of modern surface soils developed from different native rocks withinSouth China. Quat. Sci. 27, 651–652 (in Chinese).

Reis, A.P., de Almeida, L.M., da Silva, E.F., Sousa, A.J., Patinha, C., Fonseca, E.C., 2007.Assessing the geochemical inherent quality of natural soils in the Douro river basinfor grapevine cultivation using data analysis and geostatistics. Geoderma 141,370–383.

Schmidt, A., Yarnold, R., Hill, M., Ashmore, M., 2005. Magnetic susceptibility as proxy forheavy metal pollution: a site study. J. Geochem. Explor. 85, 109–117.

Statistics Bureau of Guangdong Province, Guangdong Survey Office of National Bureau ofStatistics, 2002–2011. Guangdong Statistical Yearbook (2002–2011). China StatisticsPress, Beijing.

Statistics Bureau of Guangdong Province, Office for the Population Census of GuangdongProvince, 2012. Tabulation on the 2010 Population Census of Guangdong Province.China Statistics Press, Beijing (in Chinese).

Thompson, R., Oldfield, F., 1986. Environmental Magnetism. Allen & Unwin, London.Verosub, K.L., Roberts, A.P., 1995. Environmental magnetism — past, present, and future.

J. Geophys. Res. Solid Earth 100, 2175–2192.Wang, X.S., Qin, Y., 2005. Correlation between magnetic susceptibility and heavy metals

in urban topsoil: a case study from the city of Xuzhou, China. Environ. Geol. 49,10–18.

Wang, L., Liu, D.S., Lu, H.Y., 2000. Magnetic susceptibility properties of polluted soils. Chin.Sci. Bull. 45, 1723–1726.

Wang, D.L., Lin, W.F., Yang, X.Q., Zhai, W.D., Dai, M.H., Chen, C.T.A., 2012. Occurrences ofdissolved trace metals (Cu, Cd, and Mn) in the Pearl River Estuary (China), a largeriver–groundwater–estuary system. Cont. Shelf Res. 50–51, 54–63.

Weng, Q., 2007. A historical perspective of river basin management in the Pearl RiverDelta of China. J. Environ. Manag. 85, 1048–1062.

Wong, C.S.C., Li, X.D., Zhang, G., Qi, S.H., Peng, X.Z., 2003. Atmospheric deposition of heavymetals in the Pearl River Delta, China. Atmos. Environ. 37, 767–776.

Yan, H.T., Hu, S.Y., Blaha, U., Rosler, W., Duan, X.M., Appel, E., 2011. Paddy soil— a suitabletarget for monitoring heavy metal pollution by magnetic proxies. J. Appl. Geophys.75, 211–219.

Zhang, C.X., Paterson, G.A., Liu, Q.S., 2012a. A new mechanism for the magnetic enhance-ment of hematite during heating: the role of clay minerals. Stud. Geophys. Geod. 56,845–860.

Zhang, W.G., Dong, C.Y., Ye, L.P., Ma, H.L., Yu, L.Z., 2012b. Magnetic properties of coastalloess on the Midao islands, northern China: implications for provenance andweathering intensity. Palaeogeogr. Palaeoclimatol. 333, 160–167.

Zhang, C.X., Appel, E., Qiao, Q.Q., 2013. Heavy metal pollution in farmland irrigated withriver water near a steel plant-magnetic and geochemical signature. Geophys. J. Int.192, 963–974.