direct estimates of pedogenic magnetite as a tool to

Direct estimates of pedogenic magnetite as a tool to reconstruct

past climates from buried soils

Christoph E. Geiss,1 Ramon Egli,2 and C. William Zanner3

Received 4 March 2008; revised 4 July 2008; accepted 21 August 2008; published 7 November 2008.

[1] Variations in magnetic properties of buried soils can be used to reconstruct pastclimatic conditions during paleosol formation. Most methods, however, are based oncomparisons between the magnetically enriched upper soil horizons and the magneticallyunaltered parent material. In thin loess-paleosol sequences such a comparison can beproblematic because all horizons, soil and underlying loess, may be affected to varyingdegrees by pedogenesis. We propose two direct estimates of pedogenic magnetite based onthe analysis of anhysteretic remanent magnetization ratios (cARM/isothermal remanentmagnetization) and coercivity distributions. These estimates are independent of anyinformation regarding the parent material and are possible if pedogenic minerals havesimilar magnetic properties throughout the study region. This condition seems to be metthroughout the Midwestern United States and a few loessic soils elsewhere. Theremanence-carrying part of pedogenic magnetite is composed of single-domain particleswith consistent, well-constrained magnetic properties. These particles are extremely welldispersed in the soil matrix as indicated by the absence of noticeable magnetostaticinteraction effects. Our analyses of over 70 modern loessic soil profiles demonstrate thatthe abundance of pedogenic magnetite correlates well with modern climate and that themethod is suited for reconstruction of past climates from paleosols.

Citation: Geiss, C. E., R. Egli, and C. W. Zanner (2008), Direct estimates of pedogenic magnetite as a tool to reconstruct past

climates from buried soils, J. Geophys. Res., 113, B11102, doi:10.1029/2008JB005669.

1. Introduction

[2] Many modern and buried soils are characterized by anincreased abundance of ferrimagnetic minerals (most likelymaghemite) in their upper soil horizons. This magneticenhancement of the topsoil can be detected through simplerock-magnetic measurements, and the magnetic propertiesof soils have long been used to quantify soil development.Early studies [e.g., Heller and Liu, 1986; Kukla et al., 1988]focused on changes in magnetic susceptibility (c) to recon-struct variations in climate and correlate the Chinese loess-paleosol sequence with the marine isotope record. Later,Maher and coworkers [Maher and Hu, 2006; Maher andThompson, 1992, 1995; Maher et al., 1994] studied thedifferences in susceptibility between the topsoil and theunaltered parent material. They called this difference ped-ogenic susceptibility (cped = cB�horizon � cparent material) andused it to quantify past variations in rainfall across theChinese loess plateau. Similar reconstructions of past cli-mates for the Chinese loess plateau include the work of Hanet al. [1996], who analyzed the magnetic susceptibility of

modern soils across China to reconstruct paleoclimaticvariations at Louchuan for the past 130 ka, and Heller etal. [1993], who used a combination of geochemical andmagnetic methods.[3] The use of soil magnetic properties as a paleoclimate

proxy [e.g., Heller and Evans, 1995; Maher, 1998] seemsjustified because the observed changes in soil magneticproperties are in most cases due to the neoformation of fine-grained, single-domain (SD) or even finer, superparamag-netic (SP) ferrimagnetic particles [e.g., Geiss and Zanner,2007; Maher, 1986; Mishima et al., 2001]. Though it ispossible to produce such particles in the laboratory or bysubjecting soils to a series of repeated wetting and dryingcycles [Le Borgne, 1960; Maher and Taylor, 1988; Taylor etal., 1987], it is very likely that the formation of mostpedogenic ferrimagnetic material is aided by the presenceof iron-reducing bacteria [Dearing et al., 2001; Guyodo etal., 2006;Maher and Thompson, 1999], whose activities arehighly dependent on available soil moisture, organic mattercontent, and, to a lesser degree, temperature.[4] While pedogenic susceptibility (cped) appears to be a

suitable paleorainfall proxy for the Chinese loess plateau,Geiss and Zanner [2007] found a poor relationship betweencped and mean annual precipitation for approximately 75modern loessic soils from the Midwestern United States,which they attribute to the variability of the underlyingparent material [e.g., Muhs et al., 2008] or recent additionsof Eolian dust during the Holocene and their incorporationinto the soil profile [Jacobs and Mason, 2007; Mason and

JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 113, B11102, doi:10.1029/2008JB005669, 2008ClickHere

for

FullArticle

1Department of Physics, Trinity College, Hartford, Connecticut, USA.2Department of Earth and Environmental Sciences, Ludwig-Maximi-

lians-Universitat, Munich, Germany.3Department of Soil, Water, and Climate, University of Minnesota,

St. Paul, Minnesota, USA.

Copyright 2008 by the American Geophysical Union.0148-0227/08/2008JB005669$09.00

B11102 1 of 15

http://dx.doi.org/10.1029/2008JB005669

Jacobs, 1998]. Geiss and Zanner [2006] suggested thatsimple ratios of magnetic parameters (e.g., ctopsoil/cparent

material) are better suited to compensate for parent materialheterogeneities. Since a large fraction of ferromagnetic andremanence-carrying pedogenic magnetic minerals are in theSD grain size range, Geiss and Zanner [2007] also proposedmagnetic proxies that are highly selective to this grain sizefraction of the pedogenic component, such as anhystereticremanent magnetization (ARM) or its normalized equiva-lent, susceptibility of ARM (cARM).[5] All these techniques rely on a comparison between

the magnetically enhanced soil horizons and the (presum-ably) unaltered parent material. Such an approach is prob-lematic if buried soils are separated by only a thin layer ofloess, which may be affected by pedogenesis, or if smalldust additions have been incorporated into a soil horizon. Inthis study we explore the use of two-component mixingmodels [e.g., Dunlop, 2002; Evans and Heller, 1994;Mishima et al., 2001] to quantify the abundance of pedo-genic magnetite in a soil sample without comparison to itsoriginal parent material. Such an approach is possiblebecause for many loessic soils the pedogenic magneticcomponent appears to have very similar magnetic properties[Evans and Heller, 1994; Geiss and Zanner, 2006]. Here wefocus on the use of simple remanence ratios and thecoercivity analysis of magnetization curves [Egli, 2004a;Geiss and Zanner, 2006; Robertson and France, 1994;Stockhausen, 1998] to quantify the abundance of pedogeni-cally produced ferrimagnetic minerals and their use as aclimate proxy.

2. Methods

2.1. Site Selection and Magnetic Analyses

[6] All soil profiles presented in this study have formed inPeoria loess and were sampled from well-drained, stableupland positions (Figure 1). To avoid anthropogenic influ-ences we sampled old prairie cemeteries that have remainedunaffected by agricultural activity for at least a century. Thetop 5 cm of the soil profile were not used for analysesbecause of possible contamination through fly ash fromcoal-burning power plants. Soil profiles were sampled usinga hydraulic soil probe. The 3 inch diameter cores weredescribed in the field using standard soil terminology [SoilSurvey Division Staff, 1993] and subsampled into smallplastic bags. The sampling interval was 5 cm for the mag-netically enhanced portion of the soil profile and 10 cmthereafter. All samples were air-dried in the laboratory, gentlycrushed by hand, passed through a 2 mm sieve to removeroot fragments (no gravel was present in any of the samples)and filled into weakly diamagnetic plastic boxes of 5.3 cm3

volume. Sample masses ranged between 5 and 7 g. Low-field magnetic susceptibility (c) was measured using aAGICO Kappabridge KLY 4S.[7] Anhysteretic remanent magnetization (ARM) was

acquired using a Magnon International AFD300 or a D-Tech2000 alternating-field demagnetizer. For both instrumentsthe peak AF field was 100 mTwith a bias fieldHbias = 50 mT.Susceptibility of ARM (cARM) was calculated by dividingthe mass-normalized ARM values by the bias field (Hbias)applied during ARM acquisition (cARM = ARM/Hbias).Isothermal remanent magnetization (IRM) was acquired

through three (to compensate for viscous effects duringIRM acquisition in a pulsed field) pulses of 100 mT using aASC-Scientific pulse magnetizer. All remanence parameterswere measured either in a cryogenic (2G–Model 760) or aspinner (AGICO, JR-6) magnetometer.[8] Coercivity distributions for most samples (except for

samples shown in Figures 2–4) were calculated fromalternating field (AF) demagnetization curves of an IRMacquired in a pulsed magnetic field of 2.8 T. StepwiseAF-demagnetization of the IRM up to a peak field of 300 mTwas performed with a Magnon International AFD300 alter-nating field demagnetizer. The data were fitted using twocumulative lognormal distributions as described by Geissand Zanner [2006] to model the coercivity distributions ofpedogenic magnetite, a second component which representsthe magnetic contribution of the parent material, and allhigh-coercivity minerals with switching fields <300 mT.The model function of each coercivity distribution is con-trolled by three parameters: (1) the magnitude Mcorresponding to the total magnetization of the specificcomponent, (2) the median coercivity Bh, and (3) thedispersion parameter Dp corresponding to the width of thedistribution on a log10 scale. A total of six parameters areneeded to model a magnetization curve according to thismodel without a priori assumptions (unconstrained fit). Onthe basis of numerous unconstrained analyses, Geiss andZanner [2006] noted that the pedogenic component in themagnetically enhanced horizons of loessic soils in theMidwestern United States is characterized by very consis-tent parameters given by Bh � 20 mT and Dp � 0.3, whilethe detrital component was rather heterogeneous. A differentprocedure [Egli, 2003] was used for the coercivity analysisof very detailed AF demagnetization curves of ARM andIRM shown in Figure 2. The demagnetization curves areobtained from the average of three identical experimentswhere a remanent magnetization (ARM or IRM) was given,and the sample was then stepwise demagnetized up to200 mT in a total of 105 steps, equally spaced on alogarithmic scale. To ensure exact reproducibility of theexperiments and avoid any viscous effect, AF demagneti-zation was started 3 min after the acquisition of a rema-nence, and a 10 s waiting time occurred between each AFdemagnetization step and measurement with a 2G cryogen-ic magnetometer. The high-precision experiments allowedus to model the AF demagnetization curves using a gener-alization of the lognormal distribution, called SGG [Egli,2003] that accommodates for the asymmetry of coercivitydistributions studied by Egli [2004b] The shape of a SGGfunction is controlled by an additional skewness parameter,so that a total of four parameters were used to modeleach coercivity distribution. This approach gives Bh = 18mT,Dp = 0.3 for the ARM of pedogenic magnetite in a Chinesepaleosol [Egli, 2004a], and Bh = 21 mT, Dp = 0.32 for thesample shown in Figure 2. The model parameters forpedogenic magnetite are consistent throughout the entireset of soil samples when analyzed with an unconstrainedfit, independently of the method used for coercivityanalysis. Since unconstrained fits require a large numberof precisely measured demagnetization steps, one can utilizethe constancy of the model parameters for the pedogeniccomponent, and constrain the fit algorithm to predeterminedvalues of Bh and Dp, while freely adjusting M and the

B11102 GEISS ET AL.: DIRECT ESTIMATES OF PEDOGENIC MAGNETITE

2 of 15

B11102

parameters of the other component. In this study we usedBh = 1.3 and Dp = 0.3, for the constrained analyses.[9] In this study we use several magnetic parameters to

quantify magnetic enhancement of soils. They are listed inTable 1 and can be calculated from the measurementsdescribed above.

2.2. Characterization of SD Pedogenic Magnetite

[10] In contrast to the parent material, pedogenic magne-tite particles that are blocked at room temperature aremostly single domain. Unlike SD magnetic particles pro-duced by magnetotactic bacteria, the size and shape ofpedogenic magnetite particles is not well constrained. Coer-civity analysis has been successfully employed to character-ize the SD fraction of pedogenic magnetite in loessic soils[Egli, 2004a; Evans and Heller, 1994; Geiss and Zanner,2006]. In all topsoil and paleosol samples investigated untilnow, the coercivity distribution of the pedogenic componentwas found to have strongly consistent properties Bh = 20 mT,and Dp = 0.3 as discussed in the previous section. Throughthe analysis of AF demagnetization curves of ARM and IRM(Figure 2) it is possible to calculate the ratio cARM/IRM(ARM ratio) for the pedogenic component [Egli, 2004a].The ARM ratio is strongly grain size dependent and reachesits maximum for particles whose size coincides with theupper limit for coherent moment reversal. This limit hasbeen calculated for magnetite and corresponds to a graindiameter of d = 5 nm for equidimensional particles [Newelland Merrill, 1999]. Between this limit and the SP/SDthreshold of 20 nm (calculated for magnetite particles with

a microcoercivity of 80 mT), the ARM ratio is /d2, where dis the grain diameter [Egli and Lowrie, 2002]. Using theexpression for ARM of Egli and Lowrie [2002], cARM/IRM =1� 3� 10�3 m/A for noninteracting magnetite particles. Forcomparison, the pedogenic component is characterized bycARM/IRM = 0.8 � 2 � 10�3 m/A.[11] Methods for pedogenic magnetite quantification

based on room temperature experiments rely on the SDfraction, and the constancy of the corresponding magneticproperties is of primary importance. ARM is a goodcandidate for such methods, since it is naturally selectivetoward the SD fraction; however, it is also extremelysensitive to magnetostatic interactions [Egli, 2006b]. Effec-tive volume concentrations as low as 1% are sufficient toreduce the ARM by a factor of 2. Therefore, it is importantto establish if pedogenic particles are magnetically interact-ing, and if the interaction effects are eventually the same inall soils. Quantitative methods for the analysis of interac-tions in SD particles have been developed recently [Chen etal., 2007; Egli, 2006a]. Therefore, we present here a firstevaluation of interactions in pedogenic magnetite, based ontwo samples, STP 18 and STP 70, of a soil developed onglacial till in St. Paul, Minnesota (Waukegan silt loam). Thetwo samples were taken from the soil profile at depths of18 cm and 70 cm, respectively. These two depths correspondto horizons of maximum magnetic enhancement (18 cm) andno enhancement (70 cm). The coercivity distributions calcu-lated from AF demagnetization curves of ARM and IRM areshown in Figure 2 for the enhanced. Unlike loessic soils, the

Figure 1. Location of all soil profiles analyzed for this study. All sites are located in loess, well drained,undisturbed by recent human activity, and located in stable upland positions.


3 of 15

B11102

parent material was found to contain coarse MD magnetiteand the background material is characterized by a coercivitydistribution that is unfortunately similar to that of pedogenicmagnetite. Therefore, component analysis on the bulk sam-ples does not provide reliable data in this case. In order tosolve this problem, large magnetite particles have beenextracted from the finely dispersed powder samples using astrong permanent magnet. The magnetic extraction has beencarried out until the remaining soil material did not show anyfurther decrease in magnetic susceptibility. The second pairof curves in Figure 2 shows the coercivity distributionsmeasured after magnetic extraction. While ARM remainedrelatively unaffected by the treatment, IRM was reduced byalmost a factor 2. Since ARM is strongly selective to SDparticles, we conclude that we were incapable of extractingthe SD fraction. This is probably due to the fact that finemagnetite particles are strongly bonded to large, nonmagneticminerals (probably clays). The extracted material has hyster-esis and low-temperature properties consistent with those ofMD magnetite.[12] Coercivity analysis performed after magnetic extrac-

tion is consistent with previous results obtained on loessicsamples, whereby cARM/IRM = 1 � 10�3 m3kg�1. First-order reversal curves (FORC) have been measured on bothsamples using a Micromag Alternating Force Magnetometer(AGM). Because of the weakness of the samples, nineidentical measurements have been stacked. Data have beenanalyzed using the Forcobello Matlab code [Winklhofer andZimanyi, 2006], whereby a smoothing factor SF = 3 waschosen to preserve resolution (Figure 3). The FORC dia-grams of the two samples show clear differences that can beinterpreted as the overlap of three distinct components: (1) apositive ridge along Hc = 0, which is related to reversiblemagnetization changes [Pike, 2003], not of further interesthere; (2) a FORC distribution characterized by contour linesthat are spreading out toward Hc = 0, which is characteristicfor coarse magnetite in the PSD-MD range [e.g., Roberts,2000, Figure 7]; and (3) a narrow FORC distribution con-centrated along the Hc axis [e.g., Roberts, 2000, Figure 3].Component 2 is particularly evident in the 70 cm sample,while component 3 occurs only in the 18 cm sample. Thesuperposition of the latter two components was analyzedquantitatively on a Hu profile of the FORC diagram(Figure 4a), obtained by integrating the FORC functionwith respect to Hc, between Hc = 5 mT and Hc = 27 mT(dashed lines in Figure 3). Integration was necessary toincrease the signal-to-noise ratio, especially for the 70 cmsample. Component 2 is clearly represented by a broaddistribution of Hu fields, which is identical in the twosamples. We interpret this component as the signature oflithogenic magnetic particles from the soil parent material.Superimposed to this contribution, a narrow distribution ofHu fields can be clearly seen in the 18 cm sample, althoughfaintly recognizable in the 70 cm sample as well. Since thislatter distribution is an unquestionable signature of SDparticles that occurs with highest concentration in themagnetically enhanced horizon, we can directly interpret itas a distribution of interaction fields acting between thepedogenic magnetite particles. The narrowness of the dis-tribution already suggests that magnetostatic interactions, ifpresent, are weak. Egli [2006a] calculated the theoreticalFORC distribution of weakly interacting SD particles and

Figure 2. Coercivity analysis of the 18 cm soil samplefrom St. Paul, obtained from alternating field (AF)demagnetization curves of (a) anhysteretic remanentmagnetization (ARM) susceptibility and (b) isothermalremanent magnetization (IRM). Dots represent the coerciv-ity distribution calculated by taking the first derivative ofthe demagnetization curve according to Egli [2003],whereby each dot represents a demagnetization step. Inboth plots the upper coercivity distribution corresponds tothe bulk sample, and the lower curve was measured on thesame sample after coarse multidomain (MD) magnetite wasremoved with a hand magnet. The dashed curves are modelfunctions used to fit the coercivity distribution, whereby thefit is shown as a solid curve. Each model function representsthe coercivity distribution of a magnetic component. Thelow-coercivity component is interpreted as arising frompedogenic magnetite or maghemite particles in the single-domain (SD) range, while the high-coercivity component isprobably hematite. The coercivity distribution of MDmagnetite from the parent material is similar to that of thepedogenic component and could not be resolved bycoercivity analysis. The magnetic contribution of MDmagnetite, however, is clearly shown by the differencebetween the IRM coercivity distribution measured beforeand after removing a coarse magnetic fraction with the handmagnet. As expected, ARM is only marginally affected bythe removal of coarse magnetic particles. The area under themodel function for the pedogenic component (shaded) isequal to the contribution of this component to the totalARM and IRM, respectively. The ratio of the two givescARM/IRM = 1 � 10�3 m/A.


4 of 15

B11102

concluded that Hu profiles taken for Hc < Bh coincide withthe interaction field distribution, which can be fitted withappropriate model functions to obtain the effective packingfraction of the particles. Another source of a small spreadingof Hu profiles is related to the thermal activation of viscousparticles [Egli, 2006a], as observed in goethite samples[Roberts et al., 2006]. This source cannot be excluded inour samples, since soils can contain significant amounts offine-grained goethite. Neglecting thermal activation effectswill overestimate magnetostatic interactions, meaning thatwe can obtain only an upper limit for the interaction fielddistribution. We used the method described by Egli [2006a]and Chen et al. [2007] to fit the FORC profile with twomodel curves, one that accounts for the contribution of theparent material, which will not be interpreted further, andthe second for pedogenic magnetite. FORC data processingintroduces smoothing effects that broaden narrow FORCprofiles. This effect was taken into account by calculatingthe FORC of noninteracting, rectangular hysterons andprocessing the data with the same parameters used for thesamples. The smoothing effect related to data processingwas then subtracted from the pedogenic contribution toobtain a ‘‘smoothing-free’’ profile (Figure 4a), from whichwe obtain an upper estimate of p = 0.0059 for the effectivepacking fraction of pedogenic magnetite. We can nowcompare this limit with an independent estimate of magne-tostatic interactions obtained by comparing the theoreticalARM calculated by Egli [2006b] for interacting SD particleswith the ARM ratio measured for the pedogenic compo-nent in STP18 (Figure 4). A relatively precise estimate of

cARM/IRM for noninteracting pedogenic magnetite ormaghemite particles can be obtained by integratingequation (32) in the work of Egli and Lowrie [2002]over the SD range of the pedogenic grain size distributionestimated by Liu et al. [2007]. This operation gives an upperlimit of cARM/IRM � 1.2 � 10�3 m/A for magnetite andcARM/IRM�1.4� 10�3 m/A for maghemite, using a micro-coercivity of 80 mT (corresponding to 2 � Bh) as estimatedfrom the coercivity analysis of soil samples. The experi-mental value cARM/IRM = 1.0 � 10�3 m/A obtained forthe pedogenic coercivity component of the 18 cm sample(Figure 2) corresponds to 70–80% of the theoretical valuecalculated for noninteracting particles. Considering all theuncertainties involved in modeling the ARM and in thecoercivity analysis, this difference might simply beaccounted for by calculation errors. On the other hand,assuming the cARM/IRM estimates to be error free, andassuming that the difference between the experimental valueand the predicted value for noninteracting particles isentirely produced by magnetostatic interactions, we canobtain an upper limit for the packing fraction of theparticles. For this purpose we compare the ratio of theinteracting versus noninteracting ARM predicted by Egli[2006b] for a range of packing fractions with the value of0.7–0.8 obtained above (Figure 4b), and find that this ratiocan be explained by assuming a packing fraction between0.002 and 0.005. This value is somewhat smaller than theestimate obtained from the FORCmeasurement. Consideringthe uncertainties contained both in the measurements and inthe theories of ARM, magnetostatic interactions, and

Figure 3. First-order reversal curves (FORC) diagrams for two samples from the St. Paul soil profiletaken at (a) 70 cm depth and (b) 18 cm depth. The dashed lines represent the range considered forcalculating the Hu profile shown in Figure 4.


5 of 15

B11102

Figure 4. (a) Hu profile (dots and circles) of the FORC diagrams shown in Figure 3, obtained byintegrating the FORC distribution with respect to Hc in the range between Hc = 5 and Hc = 27 mT. Theprofile of the 18 cm sample has been fitted with a combination of two model functions (dashed curves)representing theoretical interaction field distributions as calculated by Egli [2006b]. The model,consisting of the sum of the two functions (solid curve), fits the experimental data well. The narrowmodel function is corrected for the smoothing effect introduced by data processing (small dashed curve)and interpreted as the interaction field distribution of pedogenic magnetite or maghemite with a packingfraction p = 0.0059. (b) Diagram showing the effect of magnetostatic interaction on the ARM of SDparticles as a function of the packing fraction p for two grain sizes (25 nm and 40 nm) representing thelower bound of the SD range and the size that is characterized by the highest cARM/IRM, respectively.The vertical dashed line represents the upper limit for the packing fraction of the pedogenic componentestimated in Figure 4a. The two horizontal dashed lines represent the ratio between cARM/IRM measuredon the pedogenic component (Figure 2) and the theoretical cARM/IRM predicted for noninteractingmagnetite or maghemite SD particles. The solid and dashed lines define the range of packing fractions(shaded area) that is compatible with experimental data from FORC measurements and from �arm/IRM.


6 of 15

B11102

FORC, this is a good agreement. We therefore conclude thatpedogenic magnetite is only weakly interacting, wherebythe interaction effects observed, if any, are explained by avolume concentration <0.3%. This concentration corre-

sponds to a mean distance between particles of >5 timestheir diameter, which excludes the presence of dense particleaggregates as often observed in synthetic samples of SP/SDferromagnetic particles. The interaction effects of such asmall concentration on the ARM are negligible, and in anycase much smaller than the scatter observed in soil suscep-tibility. The effect of interactions on all other magneticparameters, such as susceptibility, is even smaller. Sincethe high ARM ratio of other soils analyzed suggests thesame conclusions, we can confidently exclude that magne-tostatic interactions represent a disturbing factor when usingmagnetic parameters to estimate the concentration of pedo-genic magnetite.

3. Results and Discussion

3.1. Absolute Enhancement Parameters: PedogenicSusceptibility

[13] Figure 5 shows the correlation between pedogenicsusceptibility (cped = cB horizon � cparent material) and meanannual precipitation for modern loessic soils from theChinese loess plateau [Maher et al., 1994; Porter et al.,2001], Alaska [Sharpe, 1996], the Czech Republic [Ochesand Banerjee, 1996], various sites from the NorthernHemisphere [Maher and Thompson, 1995], and the Mid-western United States (this study and that of Geiss andZanner [2007]). The best fit calibration between cped andmean annual precipitation, established by Maher et al.[1994] from an initial set of less than a dozen modern soilsfrom the Chinese loess plateau, is also shown. cped corre-lates well with mean annual precipitation for Maher et al.’s[1994] initial data and reasonably well with other magneticdata from the Chinese loess plateau [Porter et al., 2001] andthe Russian steppe [Alekseev et al., 2003], though theresulting slope of the best fit line (not shown) is a bitshallower.

Table 1. Magnetic Enhancement Parameters

Parameter Description

Absolute Enhancement Parametera

Pedogenicsusceptibility cped

cped = cB�horizon � cC�horizon

[Maher et al., 1994]

Relative Enhancement Parametersb

cenh/cparent quantifies relative change of all magneticminerals (dominated by changes inferrimagnetic minerals)

IRMenh/IRMparent quantifies relative change of remanence-carrying minerals (dominated by changesin ferrimagnetic minerals)

ARMenh/ARMparent similar to IRMenh/IRMparent but heavilybiased toward fine-grained (single-domain)ferrimagnetic minerals

ARM/IRMenh/ARM/IRMparent

quantifies changes in magnetic grain size,especially abundance of fine (SD)magnetic particles

Direct Estimates of the Pedogenic Componentc

cARM/IRM, or ARM/IRM(of enhanced horizon)

can be directly linked to abundance ofpedogenic component via two-componentmixing model [Geiss and Zanner, 2006].

IRMped/IRMtotal estimates the relative abundance of thepedogenic component through theanalysis of magnetic coercivity distributions

aQuantifies the absolute difference between the magnetically enhancedsoil horizons and the unaltered parent material.

bGive the relative change between the magnetically enhanced soilhorizons and the parent material.

cEstimate the abundance of pedogenic magnetite from one samplewithout comparison to parent material.

Figure 5. Semilogarithmic plot of pedogenic susceptibility cped = cB�horizon � cparent with meanannual precipitation for modern loessic soils. Solid line is the best fit line given by Maher et al. [1994] todescribe the relationship between cped and mean annual precipitation for modern soils from the Chineseloess plateau.


7 of 15

B11102

[14] The poor fit with other Northern Hemisphere loessdata [Maher and Thompson, 1995] may be due to regionaldifferences in the degree of magnetic enhancement, thoughthe slope of a best fit line through these considerablyscattered data would be rather shallow (decreasing its valueas a paleoprecipitation proxy). Our data from theMidwesternUnited States fall below all other previously published sites,and any correlation between cped and mean annual precip-itation is rather poor (r2 = 0.28, n = 72). Our compilation ofpedogenic susceptibility data shows that cped may be usefulfor regional climate reconstructions as long as regionaltransfer functions have been established. The large scatterin the data may also be caused by climatic influences otherthan precipitation as suggested by Han et al. [1996].

3.2. Relative Enhancement Parameters

[15] Pedogenic susceptibility, defined as the difference insusceptibility between the maximum enhanced horizon andthe background, is an example of absolute enhancementparameter. On the other hand, a relative enhancementparameter might be defined as the ratio between a givenmagnetic parameter measured for the maximally enhancedhorizon and at a depth where no enhancement can berecognized. Accordingly, a relative susceptibility enhance-ment would be defined as cenhanced/cparent. Absolute en-hancement parameters are most effective in describing aformation process for pedogenic minerals that is completelyindependent from the magnetic minerals present in the parentmaterial. A relative enhancement parameter, on the otherhand, works under the assumption that the concentration of

pedogenic magnetite is proportional with the magnetic min-erals of the parent materials.[16] The correlations between various relative magnetic

enhancement parameters and modern mean annual precip-itation for approximately 80 modern loessic soil profilesfrom the Midwestern United States are shown in Figure 6.Magnetic susceptibility (c, Figure 6a) and isothermal rem-anent magnetization (IRM, Figure 6b) represent the totalmagnetic signal produced by a wide range of magneticparticles and the correlation between these magnetic param-eters and precipitation is rather poor (r2 = 0.43 for c and r2 =0.34 for IRM). Anhysteretic remanent magnetization(ARM, Figure 6c) is heavily biased toward fine single-domain (SD) grains. Since the in situ production of suchsmall magnetic particles is the cause of magnetic enhance-ment in these soil profiles and the occurrence of SDparticles in the parent loess is negligible, an ARM-basedmagnetic enhancement parameter yields slightly betterresults (r2 = 0.53). Results improve further if the concen-tration independent ratio ARM/IRM is used to calculatemagnetic enhancement (r2 = 0.59, Figure 6d).[17] All the relative enhancement parameters shown in

Figure 6 outperform pedogenic susceptibility cped for oursites in the Midwestern United States (Figure 5). The poorperformance of cped for Midwestern soils may be due to (1)variations in the underlying parent material and (2) additionaldust input during the Holocene. Peoria loess, which is theparent material for our soil profiles, is relatively homoge-nous, but various studies [e.g.,Mason et al., 1994;Muhs and

Figure 6. (a–d) Scatterplots of relative enhancement parameters and mean annual precipitation.


8 of 15

B11102

Bettis, 2000; Muhs et al., 2008; Winspear and Pye, 1995]have identified several dust sources and transport paths.Geiss and Zanner [2007] have found significant variationsin the magnetic properties of unaltered Peoria loess. Thesevariations are at least partially due to the variability of iron-bearing magnetic minerals in the underlying parent mate-rial, which influences the degree of magnetic enhancement[Dearing et al., 1996]. Furthermore, Jacobs and Mason[2005, 2007] have shown that additions of dust during theHolocene have significantly modified many soil profiles onthe Great Plains. Given the variable sources of Peoria loess,these later additions may have different magnetic propertiesthan Pleistocene Peoria loess. Therefore, magnetic proxiesthat rely on a comparison between topsoil and parentmaterial (i.e., all the ones shown in Figures 5 and 6) maybe seriously affected by such additions of loess because theunderlying pedogenic model of downward soil developmentis now invalidated. Such later dust additions and theresulting upward growth of soils in stable upland positions(the preferred landscape position for our and similar studies)likely add to the scatter observed in Figures 5 and 6. Wetherefore interpret the partial success of relative enhance-ment parameters as due to their partial ability to account forsite-to-site variations of the parent material. However, depthvariations due for example to weathering, and later dustinput, are not accounted by relative enhancement parame-ters either, and this represent the limit that prevents a furtherimprovement of the correlation with climate.[18] Upward soil growth and the eventual modification or

loss of magnetic minerals in the lower parts of the soilprofile represent strong arguments in favor of a direct

estimate of pedogenic magnetite concentration that will bediscussed in the remainder of this paper.

3.3. Direct Estimates of Pedogenic Fraction From OneSample

[19] Geiss and Zanner [2006] as well as Evans and Heller[1994] have shown that the pedogenic component of loessicsoils from the Midwestern United States has remarkablyhomogenous magnetic properties. Knowledge of thesemagnetic properties allows us to directly estimate theamount of pedogenic magnetite present in a soil samplewithout the need for a comparison with the unaltered parentmaterial.3.3.1. ARM/IRM[20] Geiss and Zanner [2006] presented a simple linear

mixing model to estimate the abundance of pedogenicmagnetite (fped) from ARM/IRM ratios. Since ARM/IRM(or cARM/IRM) depends directly on fped it is not necessaryto calculate fped explicitly, and the average cARM/IRM of themagnetically enhanced soil horizons can be used directly asa paleoprecipitation proxy. Figure 7 shows a correlationbetween average cARM/IRM for the magnetically enhancedsoil horizons with mean annual precipitation for loessicsoils in the Midwestern United States. The correlation withmean annual precipitation is rather good (r2 = 0.70, n = 76).Furthermore, the scatter affecting all relative enhancementparameters at the wet end of our data is now greatlyreduced.[21] Aside from the good correlation with mean annual

precipitation the method is an attractive parameter becauseit requires relatively few measurements. At least in theory it

Figure 7. Average ARM ratio (cARM/IRM) of the magnetically enhanced soil horizon versus meanannual precipitation. Since cARM/IRM depends directly on the relative abundance of pedogenicmagnetite, it can be used to estimate the degree of magnetic enhancement and, if measured for buriedsoils, to estimate past climatic conditions.


9 of 15

B11102

is not necessary to measure an entire soil profile, but onecan focus solely of the modern or buried soils, which oneshould be able to identify through a careful soil profiledescription during field sampling. In addition, the measure-ment of only two remanence parameters can be carried outextremely rapidly, even when using spinner magnetometerswith characteristic measurement times of a minute or so.Using cARM/IRM as a paleoprecipitation proxy will alloweven small laboratories to analyze a large number of sites toobtain regional climate estimates. On the downside, thereduction of an entire soil profile into one ratio results in alarge loss of information, and there is little hope to extractany secondary climatic information from the data. UsingIRM acquired in a magnetic field of 100 mT rather thansaturation IRM (SIRM) to normalize for changes in theconcentration of magnetic minerals ensures that ARM andIRM are affected by the same coercivity components andminimizes the effects of mineralogical changes on thecARM/IRM proxy. Unfortunately, most published studiespresent ARM/SIRM ratios which makes comparisons be-tween sites difficult unless IRM acquisition curves are alsopublished. A comparison of our enhancement data with datafrom the Russian steppe [Alekseev et al., 2003], Alaska[Sharpe, 1996], and the Czech Republic [Oches andBanerjee, 1996] show that loessic soils from the Midwest-ern United States appear to be less magnetically enhancedthan loessic soils elsewhere, stressing the need for regionaltransfer functions between magnetic proxies and climate.3.3.2. Analysis of Magnetic Coercivity Distributions[22] The initial analysis of magnetization curves of Mid-

western soils showed that the pedogenic magnetic compo-nent is characterized by a well-constrained coercivitydistribution [Geiss and Zanner, 2006]. Figure 8 is anextension of this original data set and shows that this holdstrue for a variety of modern loessic soils from Alaska andIllinois, soils that formed in glacial outwash (J. A. TauerPrairie State Natural Area, MN, Dickman sandy loam), andeven some hydric soils that formed in extremely stony till in

Connecticut (Trinity College Field Station, Ashford, CT,Ridgebury series). The pedogenic component of all soils ischaracterized by a very narrow range of distribution param-eters Bh and Dp. Analyzing the coercivity distributions ofpaleosols using CLGs or more complicated generalizeddistribution functions [Egli, 2004a] allows us to quantita-tively estimate the amount of pedogenic magnetite directlyfrom one sample, without the need for a ‘‘background’’estimate. It is possible to use unconstrained CLG fits, butoften better results are obtained by fitting the data to apredetermined pedogenic component and a freely adjustablebackground component [Geiss and Zanner, 2006]. Figure 9shows the results of such an analysis for 15 soil profilesfrom the Midwestern United States using a predeterminedpedogenic component (Bh = 1.3, Dp = 0.3). The degree ofcorrelation with mean annual precipitation is similar toARM/IRM (r2 = 0.74, n = 15).[23] Estimates of fped using either ARM/IRM ratios or

coercivity distributions yield similar results. The muchgreater time commitment required to measure a magneticcoercivity distribution at high resolution may be rewardedby additional climatic information, such as temperature, thatcan be gained from the data. The scatter that is observed inthe magnetic properties of the pedogenic component may bejust that, or it may prove to yield additional climaticinformation. On the other hand, loessic soils from Alaskadisplay consistently lower Bh values than samples from theMidwestern United States. In the Midwestern sites, thescatter in the data is most pronounced toward the humidend of our sample area where sites follow the roughly north-south trending Iowa loess bluffs along the Missouri river.The variations in present-day mean annual temperatures aremore pronounced along this part of the transect than forother sample sites. This suggests that other climate variablesmay well influence soil magnetic properties in a systematicway, and the analysis of magnetic coercivity distributionsmay provide a means to extract these variables from soils.Unfortunately, at present, our data set does not span a

Figure 8. Logarithm of the median coercivity (log Bh) and distribution width (Dp) for the two magneticcomponents identified in magnetically enhanced soil horizons and parent material for soil profiles fromthe Midwestern United States, Alaska, and Connecticut.


10 of 15

B11102

significant enough combined precipitation and temperaturerange to test this hypothesis.[24] The direct estimates of pedogenic magnetite as out-

lined above avoid some of the shortfalls of previousmethods. Since they do not require the identification andanalysis of magnetically unaltered parent material, they canbe successful when analyzing thin loess-paleosol sequencesor soils that aggraded upward through additions of wind-blown dust that were not significant enough to interrupt soildevelopment. However, the method may run into difficultiesif later additions of dust already contain a significantamount of pedogenic magnetite, which might lead to anoverestimate of the magnetic component formed in situthrough pedogenic processes.

3.4. Application to a Welded Soil Profile

[25] Figure 10 shows some of our magnetic analyses forNorth Lutheran Cemetery in central Nebraska (40.421�N,�97.426�W). The profile contains the modern soil, a typicalprairie soil (Crete silty clay loam), and a buried soil between55 cm and 150 cm depth, identifiable through changes insoil color and structure and otherwise incorporated into themodern soil. The lower part of this buried soil (90–120 cm,interpreted as a B horizon with former E horizon character-istics) is magnetically enhanced leading to increased valuesin all concentration-dependent magnetic parameters (cshown in Figure 10a). ARM ratios (Figure 10b) show ageneral increase in fine-grained SD material throughout theupper 150 cm of the soil profile indicating that magneticenhancement processes were active throughout the deposi-tion of approximately 50 cm of Holocene loess. An uncon-strained analysis of magnetic coercivity distributions(Figure 10d) shows that the magnetically enhanced buriedsoil horizon is enriched in a pedogenic component (solidblack curves) similar to the ones found in the magnetically

enhanced modern soil horizon. This component is absentbetween 50–80 cm depth (Bt2 horizon), the assumed top ofthe buried soil. Despite possible pedogenic overprints thesimilarities of the pedogenic components in both the modernand the buried soil may allow us to interpret the magneticrecord. Figure 10c shows our analyses of magnetic coercivityspectra where the pedogenic component was constrained tolog Bh = 1.3 andDp = 0.3. The resulting abundance estimatesfor the pedogenic component suggest that the buried soilformed under slightly wetter climatic conditions. Furtheranalyses of similar, but better preserved, paleosols are cur-rently under way and will show whether it is possible toextract climatic information from such poorly separatedpaleosol horizons.

3.5. Inherited Pedogenic Magnetic Components

[26] Holocene (Bignell) loess in Nebraska and elsewherein the Midwestern United States is often darker in color,possibly because it has been derived from a mixture offloodplain deposits and older reworked loess and soil[Mason and Kuzila, 2000]. We analyzed the magneticcoercivity distributions of two samples of Bignell andPeoria loess to test for the presence of a pedogenic compo-nent in Bignell loess. The samples were supplied to us byMark Kuzila and are from the Mirdan Canal section incentral Nebraska [Mason and Kuzila, 2000]. The results areshown in Figure 11. Both loess units contain a magneticcomponent that could be of pedogenic origin, but theirabundance varies little between the two samples. Thiscomponent can be an intrinsic part of the parent material(as is the case for the sample analyzed in Figure 2) or isevidence for continued pedogenesis during periods of loessdeposition. In either case, the magnetic properties of bothloess units are sufficiently similar to allow for meaningful

Figure 9. Relative IRM contributions of the pedogenic magnetic component versus mean annualprecipitation.


11 of 15

B11102

direct estimates of pedogenically formed magnetite in soilprofiles that received later additions of Bignell loess.

4. Conclusions

[27] Both ARM ratios (cARM/IRM) and magnetic coer-civity distributions reflect climatic conditions during soilformation and can be used as a paleoclimate proxy. Formodern soils from the Midwestern United States the corre-lation between these two magnetic proxies and paleoclimateis better than for any other previously studied magneticproxy (cped in Figure 5 and various magnetic ratios in

Figure 6) and independent of our knowledge of the original,unaltered parent material. The latter fact allows it to beapplicable to the analysis of thin loess-paleosol sequences,where no unaltered parent material may be available.[28] Our analyses of the various magnetic enhancement

parameters show that the correlation between magneticproxies and climate is only of regional value. This maynot be a problem as long as a regional transfer function canbe established from modern soils. However, it is possiblethat these transfer functions change not only in space butalso in time, making their application to older soils difficult

Figure 10. Magnetic analyses of a soil profile obtained from North Lutheran Cemetery, Nebraska(40.421�N, �97.426�W). (a) Variations in magnetic susceptibility (c) reflect changes in the abundance ofall iron-bearing minerals but are heavily biased by changes in ferrimagnetic maghemite or magnetite.(b) The ratio of cARM/IRM serves as a proxy for the relative abundance of fine-grained (SD) magneticparticles. (c) Estimate of the relative abundance of pedogenic magnetite based on the analysis ofcoercivity distributions. (d) Unconstrained analyses of magnetic coercivity distributions for selectedsamples from North Lutheran Cemetery showing the presence of a pedogenic magnetic component (solidcurves) in the magnetically enhanced horizons as well as a magnetic background component (dashedcurves). No pedogenic component was detected in samples below 150 cm depth, and both coercivitycomponents were classified as background components. Abbreviations used in soil profile description:str, strong; mod, moderate; gran, granular; abk, angular blocky; sbk, subangular blocky; prism, prismatic;f, fine; med, medium; c, coarse; sil, silt loam; sicl, silty clay loam; h, heavy.


12 of 15

B11102

if not impossible. More research regarding the processesthat influence magnetic enhancement is clearly necessary.[29] Our study limited itself to modern soils that formed

within the last few thousand years. Some authors [e.g.,Maher and Thompson, 1999, and references therein] arguethat magnetic enhancement proceeds relatively quickly andreaches saturation within a few hundred or thousand yearsin loessic soils. This view is not shared by several other

authors [e.g., Grimley et al., 2003; Singer et al., 1992; Vidicand Lobnik, 1997; Vidic et al., 2004] whose researchsuggests that the magnetic signal builds up more slowlyover time. In light of such uncertainties, it might be best,when using a transfer function, to limit oneself to paleosolsthat are thought to have formed for similar lengths of timeas the modern soils on which the function is based.

Figure 11. Coercivity distributions and coercivity component analyses for a sample of (a) Bignell loessand (b) Peoria loess. Dashed curves show individual coercivity components, solid curve shows the sumof the two individual coercivity components, and dots show measured coercivity data.


13 of 15

B11102

[30] Despite these limitations, magnetic analyses of bur-ied soils can provide quantitative information on pastprecipitation gradients, though more work is clearly neces-sary to constrain the pathways of magnetic enhancementand their link to biologically driven processes in the uppersoil horizons.

[31] Acknowledgments. We are grateful to all landowners, funeralhomes, and land managers for access to the numerous sites acrossNebraska, Iowa, Missouri, Illinois, and Minnesota. Jim Bisbee, DanScollan, Alex Masi, and Saroj Aryal helped with field sampling andmagnetic analyses at Trinity College. Mark Kuzila (University of Nebraskaat Lincoln) generously supported this project throughout the years bysupplying a probe truck and by supplying samples from the Mirdan Canalsection. Most of C. E. G.’s work was supported through faculty and studentresearch grants from Trinity College. R. E. is grateful to Amy Chen formeasuring FORC diagrams of soil samples. Some of the magnetic measure-ments were conducted at the Institute for Rock Magnetism (IRM). The IRMis funded by the W.M. Keck Foundation, the National Science Foundation’sEarth Science Division’s Instrumentation and Facilities Program, and theUniversity of Minnesota.

ReferencesAlekseev, A. O., T. V. Alekseeva, and B. A. Maher (2003), Magneticproperties and mineralogy of iron compounds in steppe soils, EurasianSoil Sci., 36(1), 59–70.

Chen, A. P., R. Egli, and B. M. Moskowitz (2007), First-order reversalcurve (FORC) diagrams of natural and cultured biogenic magnetic par-ticles, J. Geophys. Res., 112, B08S90, doi:10.1029/2006JB004575.

Dearing, J. A., K. L. Hay, S. M. J. Baban, A. S. Huddleston, E. M. H.Wellington, and P. J. Loveland (1996), Magnetic susceptibility of soil: Anevaluation of conflicting theories using a national data set, Geophys. J.Int., 127, 728–734, doi:10.1111/j.1365-246X.1996.tb04051.x.

Dearing, J. A., J. A. Hannam, A. S. Anderson, and E. M. H. Wellington(2001), Magnetic, geochemical and DNA properties of highly magneticsoils in England, Geophys. J. Int., 144, 183–196, doi:10.1046/j.0956-540X.2000.01312.x.

Dunlop, D. J. (2002), Theory and application of the Day plot (Mrs/Ms

versus Hcr/Hc): 2. Application to data for rocks, sediments, and soils,J. Geophys. Res., 107(B3), 2057, doi:10.1029/2001JB000487.

Egli, R. (2003), Analysis of the field dependence of remanent magnetiza-tion curves, J. Geophys. Res. , 108(B2), 2081, doi:10.1029/2002JB002023.

Egli, R. (2004a), Characterization of individual rock magnetic componentsby analysis of remanence curves, 1, Unmixing natural sediments, Stud.Geophys. Geod., 48, 391 – 446, doi:10.1023/B:SGEG.0000020839.45304.6d.

Egli, R. (2004b), Characterization of individual rock-magnetic componentsby analysis of remanence curves, 2, Fundamental properties of coercivitydistributions, Phys. Chem. Earth, 29, 851–867.

Egli, R. (2006a), Theoretical aspects of dipolar interactions and their appear-ance in first-order reversal curves of thermally activated single-domainparticles, J. Geophys. Res., 111, B12S17, doi:10.1029/2006JB004567.

Egli, R. (2006b), Theoretical considerations on the anhysteretic remanentmagnetization of interacting particles with uniaxial anisotropy, J. Geo-phys. Res., 111, B12S18, doi:10.1029/2006JB004577.

Egli, R., and W. Lowrie (2002), Anhysteretic remanent magnetization offine magnetic particles, J. Geophys. Res., 107(B10), 2209, doi:10.1029/2001JB000671.

Evans, M. E., and F. Heller (1994), Magnetic enhancement and paleocli-mate: Study of a loess/paleosol couplet across the loess plateau of China,Geophys. J. Int . , 117 , 257 – 264, doi:10.1111/j .1365-246X.1994.tb03316.x.

Geiss, C. E., and C. W. Zanner (2006), How abundant is pedogenic magne-tite: Abundance and grain-size estimates for loessic soils based on rock-magnetic analyses, J. Geophys. Res., 111, B12S21, doi:10.1029/2006JB004564.

Geiss, C. E., and C. W. Zanner (2007), Sediment magnetic signature ofclimate in modern loessic soils from the Great Plains, Quat. Int., 162–163, 97–110, doi:10.1016/j.quaint.2006.10.035.

Grimley, D. A., L. R. Follmer, R. E. Hughes, and P. A. Solheid (2003),Modern, Sangamon and Yarmouth soil development in loess of ungla-ciated southwestern Illinois, Quat. Sci. Rev., 22, 225–244, doi:10.1016/S0277-3791(02)00039-2.

Guyodo, Y., T. M. LaPara, A. J. Anschutz, R. L. Penn, S. K. Banerjee, C. E.Geiss, and C. W. Zanner (2006), Rock magnetic, chemical and bacterial

community analysis of a modern soil from Nebraska, Earth Planet. Sci.Lett., 251, 168–178, doi:10.1016/j.epsl.2006.09.005.

Han, J., H. Lu, N. Wu, and Z. Guo (1996), The magnetic susceptibility ofmodern soils in China and its use for paleoclimate reconstruction, Stud.Geophys. Geod., 40, 262–275, doi:10.1007/BF02300742.

Heller, F., and M. E. Evans (1995), Loess magnetism, Rev. Geophys., 33,211–240.

Heller, F., and T. S. Liu (1986), Paleoclimatic and sedimentary history frommagnetic susceptibility of loess in China, Geophys. Res. Lett., 13, 1169–1172.

Heller, F., C. D. Shen, J. Beer, T. S. Liu, A. Bronger, M. Suter, and G. Bonani(1993), Quantitative estimates of pedogenic ferromagnetic mineral forma-tion in Chinese loess and paleoclimatic implications, Earth Planet. Sci.Lett., 114, 385–390, doi:10.1016/0012-821X (93)90038-B.

Jacobs, P. M., and J. A. Mason (2005), Impact of Holocene dust aggrada-tion on A horizon characteristics and carbon storage in loess-derivedMollisols of the Great Plains, USA, Geoderma, 125, 95 – 106,doi:10.1016/j.geoderma.2004.07.002.

Jacobs, P. M., and J. A. Mason (2007), Late Quaternary climate change,loess sedimentation, and soil profile development in the central GreatPlains: A pedosedimentary model, Geol. Soc. Am. Bull., 119(3), 462–475, doi:10.1130/B25868.1.

Kukla, G., F. Heller, X. M. Liu, T. C. Xu, T. S. Liu, and Z. S. An (1988),Pleistocene climates in China dated by magnetic susceptibility, Geology,16, 811 –814, doi:10.1130/0091-7613(1988)016<0811:PCICDB>2.3.CO;2.

Le Borgne, E. (1960), Influence du feu sur les proprietes magnetiques dusol et sur celles du schiste et du granite, Ann. Geophys., 16, 159–195.

Liu, Q., C. Deng, J. Torrent, and R. Zhu (2007), Review of recent devel-opments in mineral magnetism of the Chinese loess, Quat. Sci. Rev., 26,368–385, doi:10.1016/j.quascirev.2006.08.004.

Maher, B. A. (1986), Characterization of soils by mineral magnetic mea-surements, Phys. Earth Planet. Inter., 42, 76–92, doi:10.1016/S0031-9201(86)80010-3.

Maher, B. A. (1998), Magnetic properties of modern soils and Quaternaryloessic paleosols: Paleoclimatic implications, Palaeogeogr. Palaeoclima-tol. Palaeoecol., 137, 25–54, doi:10.1016/S0031-0182(97)00103-X.

Maher, B. A., and M. Hu (2006), A high-resolution record of Holocenerainfall variations from the western Chinese Loess Plateau: Antiphasebehaviour of the African/Indian and East Asian summer monsoons,Holocene, 16(3), 309–319, doi:10.1191/0959683606hl929rp.

Maher, B. A., and R. M. Taylor (1988), Formation of ultrafine-grainedmagnetite in soils, Nature, 336, 368–370, doi:10.1038/336368a0.

Maher, B. A., and R. Thompson (1992), Paleoclimatic significance of themineral magnetic record of the Chinese loess and paleosols, Quat. Res.,37, 155–170, doi:10.1016/0033-5894(92)90079-X.

Maher, B. A., and R. Thompson (1995), Paleorainfall reconstructions frompedogenic magnetic susceptibility variations in the Chinese loess andpaleosols, Quat. Res., 44, 383–391, doi:10.1006/qres.1995.1083.

Maher, B. A., and R. Thompson (1999), Palaeomonsoons, I: The magneticrecord of palaeoclimate in the terrestrial loess and palaeosol sequences, inQuaternary Climates, Environments and Magnetism, edited by B. A.Maher and R. Thompson, pp. 81–125, Cambridge Univ. Press, Cambridge.

Maher, B. A., R. Thompson, and L. P. Zhou (1994), Spatial and temporalreconstructions of changes in the Asian paleomonsoon: A new mineral-magnetic approach, Earth Planet. Sci. Lett., 125, 461–471, doi:10.1016/0012-821X(94)90232-1.

Mason, J. A., and P. M. Jacobs (1998), Chemical and particle-size evidencefor addition of fine dust to soils of the midwestern United States, Geology,26, 1135–1138, doi:10.1130/0091-7613(1998)026<1135:CAPSEF>2.3.CO;2.

Mason, J. A., and M. S. Kuzila (2000), Episodic Holocene loess depositionin central Nebraska, Quat. Int., 67, 119 – 131, doi:10.1016/S1040-6182(00)00013-6.

Mason, J. A., E. A. Nater, and H. C. Hobbs (1994), Transport direction ofWisconsinian loess in southeastern Minnesota, Quat. Res., 41, 44–51,doi:10.1006/qres.1994.1005.

Mishima, T., X.-M. Fang, B.-T. Pan, J.-J. Li, M. Torii, H. Fukusawa,and Y. Ono (2001), Magnetic grain-size distribution of the enhanced com-ponent in the loess-palaeosol sequences in the western Loess Plateau ofChina, Geophys. J. Int., 145(2), 499 – 504, doi:10.1046/j.1365-246x.2001.01397.x.

Muhs, D. R., and E. A. I. Bettis (2000), Geochemical variations in Peorialoess of western Iowa indicate paleowinds of midcontinental NorthAmerica during last glaciation, Quat. Res., 53, 49–61, doi:10.1006/qres.1999.2090.

Muhs, D. R., E. A. I. Bettis, J. N. Aleinikoff, J. P. McGeehin, J. Beann,G. Skipp, H. Roberts, W. C. Johnson, R. Benton, and T. A. Ager (2008),Origin and paleoclimatic significance of late Quaternary loess in Nebraska:


14 of 15

B11102

Stratigraphy, chronology, sedimentology, and geochemistry, Geol. Soc.Am. Bull., doi:10.1130/B26221.1, in press.

Newell, A. J., and R. T. Merrill (1999), Single-domain critical sizes forcoercivity and remanence, J. Geophys. Res., 104, 617–626.

Oches, E. A., and S. K. Banerjee (1996), Rock-magnetic proxies of climatechange from loess-paleosol sediments of the Czech Republic, Stud. Geo-phys. Geod., 40, 287–300, doi:10.1007/BF02300744.

Pike, C. R. (2003), First-order reversal curve diagrams and reversible mag-netization, Phys. Rev. B, 68, 104424, doi:10.1103/PhysRevB.68.104424.

Porter, S. C., B. Hallet, X. Wu, and Z. An (2001), Dependence of near-surface magnetic susceptibility on dust accumulation rate and precipita-tion on the Chinese loess plateau, Quat. Res., 55, 271–283, doi:10.1006/qres.2001.2224.

Roberts, A. P. (2000), First-order reversal curve diagrams: A new tool forcharacterizing the magnetic properties of natural samples, J. Geophys.Res., 105(B12), 28,461–28,475.

Roberts, A. P., Q. S. Liu, C. J. Rowan, L. Chang, C. Cavallo, J. Torrent,and C.-S. Horng (2006), Characterization of hematite (a-Fe2O3),goethite (a-FeOOH), greigite (Fe3S4), and pyrrhotite (Fe7S8) usingfirst-order reversal curve diagrams, J. Geophys. Res., 111, B12S35,doi:10.1029/2006JB004715.

Robertson, D. J., and D. E. France (1994), Discrimination of remanence-carrying minerals in mixtures, using isothermal remanent magnetizationacquisition curves, Phys. Earth Planet. Inter., 82, 223–234, doi:10.1016/0031-9201(94)90074-4.

Sharpe, J. L. (1996), Environmental archeology in central Alaska: A mag-neto-stratigraphic correlation of Tanana valley archeological sites withglobal climatic change, M. A. thesis, Univ. of Minn., Minneapolis.

Singer, M. J., P. Fine, K. L. Verosub, and O. A. Chadwick (1992), Timedependence of magnetic susceptibility of soil chronosequences on theCalifornia coast, Quat. Res., 37, 323 – 332, doi:10.1016/0033-5894(92)90070-Y.

Soil Survey Division Staff (1993), Soil Survey Manual, 437 pp., U.S. Gov.Print. Off., Washington, D.C.

Stockhausen, H. (1998), Some new aspects for the modeling of isothermalremanent magnetization acquisition curves by cumulative log Gaussianfunctions, Geophys. Res. Lett., 25(12), 2217–2220.

Taylor, R. M., B. A. Maher, and P. G. Self (1987), Magnetite in soils: I. Thesynthesis of single-domain and superparamagnetic magnetite, ClayMiner., 22, 411–422, doi:10.1180/claymin.1987.022.4.05.

Vidic, N., and F. Lobnik (1997), Rates of soil development of the chronose-quence in the Ljubljana basin, Slovenia, Geoderma, 76, 35 – 64,doi:10.1016/S0016-7061(96)00098-5.

Vidic, N. J., M. J. Singer, and K. L. Verosub (2004), Duration dependenceof magnetic susceptibility enhancement in the Chinese loess-paleosols ofthe past 620 ky, Palaeogeogr. Palaeoclimatol. Palaeoecol., 211, 271–288, doi:10.1016/j.palaeo.2004.05.012.

Winklhofer, M., and G. T. Zimanyi (2006), Extracting the intrinsic switch-ing field distribution in perpendicular media: A comparative analysis,J. Appl. Phys., 99, 08E710, doi:10.1063/1.2176598.

Winspear, N. R., and K. Pye (1995), Textural, geochemical and mineralo-gical evidence for the origin of Peoria loess in central and southernNebraska, USA, Earth Surf. Processes Landforms, 20, 735 – 745,doi:10.1002/esp.3290200805.

��R. Egli, Department of Earth and Environmental Sciences, Ludwig-

Maximilians-Universitat, Geophysics Theresienstrasse 41, Munich D-80333,Germany.C. E. Geiss, Department of Physics, Trinity College, 300 Summit Street,

McCook 105, Hartford, CT 06106, USA. ([email protected])C. W. Zanner, Department of Soil, Water, and Climate, University of

Minnesota, St. Paul, MN 55108, USA.


15 of 15

B11102

direct estimates of pedogenic magnetite as a tool to

Documents