Separation of pedogenic and lithogenic components of...

Geophys. J. Int. (2000) 142, 551–562

Separation of pedogenic and lithogenic components of magneticsusceptibility in the Chinese loess/palaeosol sequence as determinedby the CBD procedure and a mixing analysis

Natasa J. Vidic,1,2,3 Jeff D. TenPas,2 Kenneth L. Verosub3 and Michael J. Singer21Agronomy Department, University of L jubljana, 1111 L jubljana, Slovenia. E-mail: [email protected] of L and, Air and Water Resources, University of California, Davis, CA 95694, USA3Department of Geological Sciences, University of California, Davis, CA 95694,USA

Accepted 2000 March 20. Received 2000 March 17; in original form 1999 July 13

SUMMARYMagnetic susceptibility variations in the Chinese loess/palaeosol sequences have beenused extensively for palaeoclimatic interpretations. The magnetic signal of thesesequences must be divided into lithogenic and pedogenic components because thepalaeoclimatic record is primarily reflected in the pedogenic component. In this paperwe compare two methods for separating the pedogenic and lithogenic components ofthe magnetic susceptibility signal: the citrate-bicarbonate-dithionite (CBD) extractionprocedure, and a mixing analysis. Both methods yield good estimates of the pedogeniccomponent, especially for the palaeosols. The CBD procedure underestimates thelithogenic component and overestimates the pedogenic component. The magnitude ofthis effect is moderately high in loess layers but almost negligible in palaeosols. Themixing model overestimates the lithogenic component and underestimates the pedogeniccomponent. Both methods can be adjusted to yield better estimates of both components.The lithogenic susceptibility, as determined by either method, suggests that palaeo-climatic interpretations based only on total susceptibility will be in error and that asingle estimate of the average lithogenic susceptibility is not an accurate basis foradjusting the total susceptibility. A long-term decline in lithogenic susceptibility withdepth in the section suggests more intense or prolonged periods of weathering associatedwith the formation of the older palaeosols.The CBD procedure provides the most comprehensive information on the magnitude

of the components and magnetic mineralogy of loess and palaeosols. However, themixing analysis provides a sensitive, rapid, and easily applied alternative to the CBDprocedure. A combination of the two approaches provides the most powerful andperhaps the most accurate way of separating the magnetic susceptibility components.

Key words: China, loess, magnetic susceptibility, palaeoclimate, Quaternary

dependence between the pedogenic and lithogenic components1 INTRODUCTION

of magnetic susceptibility. We also used the two methods to

The purpose of this paper is to evaluate and compare two determine the magnetic mineralogy of the two components.

methodologies that we have developed for separating the pedo- Thick sequences of interstratified loess and palaeosols from

genic and lithogenic components of the magnetic susceptibility the Chinese loess plateau provide one of the most complete

and most sensitive terrestrial records of environmental changesignal of loess/palaeosol sequences. The focus of our attention

is the upper part (L1–L6) of Jiaodao section in the Chinese over the past 2.5 million years (Heller & Liu 1986; Liu 1988;

Kukla & An 1989; Kukla et al. 1990; An et al. 1991; Ruttercentral loess plateau (50 km north of Luochuan). The first

method involves magnetic measurements before and after 1992). The alternating couplets of loess and palaeosols in the

laterally extensive sequences are interpreted as having formedcitrate-bicarbonate-dithionite (CBD) extraction of pedogenic

iron minerals. The second method uses a mathematical in response to fluctuations in the strength of the dry, Siberian–

Mongolian (winter) monsoon, and the wet, East Asian andmixing analysis that is based on the contrast in the frequency

551© 2000 RAS

552 N. J. V idic et al.

Indian (summer) monsoons (An et al. 1991). The palaeosols that can be used as an alternative to the CBD procedure. The

are believed to have formed during periods of diminished dust mixing analysis (TenPas 1997; TenPas et al. 1999) is baseddeposition and enhanced soil formation, when the regional on the (pre-CBD) magnetic susceptibility and its frequencyclimate was dominated by a strong summer monsoon. In dependence plus values of the inherent frequency dependencecontrast, loess deposition prevailed when the regional climate of purely lithogenic and pedogenic material. These values canwas dominated by a strong winter monsoon. At all times, be estimated from pre-CBD measurements or determined fromhowever, loess deposition and pedogenesis were competing post-CBD measurements on a small subset of samples, thusprocesses, and only their relative strengths varied as the climate reducing the need for time-consuming extractions.fluctuated (Verosub et al. 1993; Fine et al. 1995).Because pedogenesis has enhanced the magnetic susceptibility

of the palaeosols compared with the loess (Zhou et al. 1990;2 MATERIALS AND METHODSMaher & Thompson 1991; Verosub et al. 1993), variations in

magnetic susceptibility of the Chinese loess/palaeosol sequences2.1 Setting and samplingoffer the potential for palaeoclimatic reconstructions (Heller

et al. 1993; Maher & Thompson 1995; Liu et al. 1995; Han The Jiaodao loess/palaeosol section is located in the centralet al. 1996). Several studies have shown that the magnetic area of the Chinese loess plateau about 50 km north of thesignal of these sequences can be divided into lithogenic and classic Luochuan locality. The mean annual precipitation ofpedogenic components (Zhou et al. 1990; Maher & Thompson Fuxian County is 600 mm. Most precipitation falls between1991; Verosub et al. 1993; Banerjee et al. 1993; Evans & Heller April and October. The mean annual temperature is 8.9 °C.1994; Fine et al. 1995). The lithogenic component is carried Seasonal temperature variations are quite high: the mean Julypredominantly by coarse-grained multidomain ferrimagnetic temperature is 23.3 °C, but the mean January temperature isgrains, while the pedogenic component is carried by fine- only −6.5 °C (oral communication with C. L. Deng, Chinesegrained superaparamagnetic to single-domain magnetic material Academy of Sciences, Beijing). The present soil temperature(Hus & Han 1992; Maher & Thompson 1992; Banerjee et al. regime is mesic, and the present soil moisture regime ustic.1993; Fine et al. 1993; Evans & Heller 1994; TenPas 1997). Our samples come from 48 m of the upper part of the sectionThe separation of the total susceptibility signal into its starting at approximately 5 m from the surface. They representpedogenic and lithogenic components has been challenging.

the lower part of the Malan loess (L1), and a large section ofSeveral studies (Maher & Thompson 1992; Banerjee et al.

the Lishi loess (S1–L6) (Kukla & An 1989). Each bulk sample1993; Beer et al. 1993; Heller et al. 1993; Forster et al. 1994)

contained at least 200 g of material. Samples of loess L1 to L6have used methods that are not accessible to most researchers,

and intercalated palaeosols were collected at 20 cm intervals,such as 10Be, or that yield average lithogenic component

while the transitions from loess to palaeosols were collected atestimates for the entire section. In contrast, our methods can

10 cm intervals.be used in most laboratories and yield estimates of the litho-

genic and pedogenic components for each sample (Fine et al.1993; Verosub et al. 1993; Fine et al. 1995; Singer et al. 1995).Our approach involves differential dissolution of secondary 2.2 Laboratory proceduresiron oxides using a citrate-bicarbonate-dithionite (CBD) treat-

Samples were air-dried and gently crushed to pass throughment as well as determination of pre- and post-CBD magnetica 2 mm sieve. We then packed 4 g of material into smallsusceptibilities and their frequency dependence. The CBD(1.9×1.9×1.6 cm) plastic boxes for susceptibility measure-procedure (Mehra & Jackson 1960; Janitzky 1986) has beenments. Another 4 g of the same sample was treated with CBDused for years by soil scientists to selectively dissolve pedogenicfollowing Janitzky (1986). The CBD extraction was done twiceFe-compounds. Our previous work (Fine et al. 1989; Fine et al.using 2 g additions of sodium dithionite each time. During1993; Fine et al. 1995; Hunt et al. 1995; Singer et al. 1995) hasthe extraction, the samples were heated to 75 °C. Followingshown that the procedure is particularly effective at removingeach extraction the sample was centrifuged, the supernatantmaghemite and any antiferromagnetic components (such aswas decanted, and the solid portion was washed with sodiumhaematite and goethite) as well as fine-grained magnetite. Sincecitrate. The supernatant from the wash was added to theother studies have shown that the lithogenic component in theoriginal supernatant and saved for determination of CBD-loess/palaeosol sequence is carried predominantly by coarse-extractable Fe. A total of 4 g of sodium dithionite was usedgrained magnetite (Hus & Han 1992; Maher & Thompsonfor each 4 g of sample, and the total reaction time was 1 hr.1992; Banerjee et al. 1993; Evans & Heller 1994), the post-The solid residue was dried at 40 °C and packed into a plasticCBD value of low-field magnetic susceptibility provides a goodbox for magnetic susceptibility measurements.first approximation of the contribution from the lithogenicHigh- (4.7 kHz) and low- (0.47 kHz) frequency magneticcomponent. However, since the lower limit of the grain size

susceptibilities were determined on pre- and post-CBD samplesdistribution of the lithogenic component is not known andusing a Bartington MS-2 magnetic susceptibility meter withsince the CBD procedure is known to remove some fine-MS-2B sensor. For both pre- and post-CBD samples, thegrained (<1 mm) synthetic magnetite (Hunt et al. 1995), therefrequency-dependent part of the susceptibility was calculatedis the possibility that the CBD procedure may lead to anas the difference between the low- and high-frequency mag-underestimate of the lithogenic component.netic susceptibility values. (Note that in this paper, the termOne goal of this paper is to estimate the degree to which‘frequency dependence of magnetic susceptibility’ refers to thethe CBD procedure removes lithogenic susceptibility fromactual change in susceptibility rather than the per cent changeloess and palaeosol samples and to calculate a correction

factor for it. We also present and discuss a mixing analysis in susceptibility.)

© 2000 RAS, GJI 142, 551–562

Separation of pedogenic and lithogenic susceptibility 553

The Fe extracted by CBD (Fed) was determined on a Jarrell magnetic susceptibilities and their frequency-dependent parts

as discussed in Section 3.2. A summary of the symbols usedAsh inductively coupled plasma spectrometer. Total Fe (Fet)was determined for a subset of Lishi samples using the plasma can be found in Table 1.spectrometer after dissolution of the samples with HF.

To examine the effects of CBD on natural samples, weapplied the stepwise CBD procedure used by Hunt et al. (1995) 3 RESULTSfor synthetic samples to two loess/palaeosol samples from the

upper part of the Jiaodao section and to 10 North American 3.1 CBDmethodologyloess samples. In this case, the samples were subjected to afour-step extraction using 1 g of sodium dithionite for each 3.1.1 Variations in magnetic measurements and ironstep. The magnetic susceptibility of the solid residue was concentrationsmeasured after each step.

Pre-CBD magnetic susceptibility (xpre ) represents the totalmagnetic susceptibility signal. Fig. 1 shows that palaeosolsdisplay significantly higher pre-CBD magnetic susceptibility2.3 Mixing analysisvalues (120–300× 10−8 m3 kg−1) and a higher per cent

Mixing analysis (TenPas 1997; TenPas et al. 1999) is a frequency dependence (fdt= 8–12 per cent) than loessmathematical technique that assumes that a quantity being (25–75×10−8m3 kg−1, 3–8 per cent, respectively) (Fig. 1).measured results from the admixture of two end members However, xpre is dominated by the CBD-extractable mag-with fixed properties. In this case, the two end members are netic susceptibility (xext) (Fig. 1), with palaeosols displayingthe lithogenic and pedogenic components. The advantage of much higher values (100–290×10−8m3 kg−1 ) than loessthe mixing analysis is that more easily measured variables in (15–50×10−8m3 kg−1 ). CBD treatment removes 80–95 pertwo sets of equations can be used to solve for variables that cent of the low-frequency susceptibility from the palaeosols,are more difficult to measure. We use the mixing analysis to but only 50–75 per cent from the loess.calculate the pedogenic and lithogenic contributions to the As discussed above, the post-CBD magnetic susceptibilitysusceptibility with the goal of reducing the number of time- (xpost) provides an estimate of the contribution from theconsuming CBD extractions that are needed to estimate these lithogenic component. Compared with xpre ,

xpost is generally

contributions. low and only moderately variable (Fig. 1). The values varyWe begin by writing the total low-frequency susceptibility between 9 and 28×10−8 m3 kg−1, decrease with the depthas the sum of the component susceptibilities: of the section, and display some higher-frequency variations

(Figs 1 and 5).xt=xl+xp , (1) CBD-extractable iron (Fed) is consistently higher in palaeo-

sols (1–2 per cent) than in loess (<1 per cent) (Fig. 1). Totalwhere xt is the total low-frequency susceptibility and

xl and

xp Fe (Fet), which has been measured only in part of the section,are the lithogenic and pedogenic low-frequency susceptibilities, shows variations that are consistent with the alternation of

respectively. Furthermore, loess–palaeosol layers as well (Fig. 1). It is higher in palaeosols

(3.5–4 per cent) than in loess (<3.5 per cent) (Fig. 1). However,xfdt

=xfdl

+xfdp

, (2)

where xfdt

is the total frequency-dependent part of thesusceptibility, and xfd

l

and xfdp

are the lithogenic and pedo- Table 1. List of symbols.genic contributions, respectively. We now assume that the

frequency-dependent part of the pedogenic susceptibility is symbol descriptiondirectly proportional to the pedogenic susceptibility itself,

xpre pre-CBD low-frequency magnetic susceptibilityand that the proportionality factor (the per cent frequencyxpost post-CBD low-frequency magnetic susceptibilitydependence fdp ) is the same for all pedogenic material. We xext CBD-extractable low-frequency magnetic susceptibilitymake the same assumption about the frequency dependencexfdpre

pre-CBD frequency dependence of magnetic susceptibilityof the lithogenic susceptibility and about its proportionalityxfdpost

post-CBD frequency dependence of magnetic susceptibilityfactor (fdl ). Hence xfdext

extractable frequency dependence of magnetic susceptibilityxt total low-field magnetic susceptibility is the same as xprex

fdt

= fdlxl+ fdp

xp . (3)

xp pedogenic magnetic susceptibility component

(mixing analysis)Provided fdl and fdp are known or can be determined, we x

l lithogenic magnetic susceptibility componentcan use eqs (1) and (3) to solve for xl and

xp using measurements (mixing analysis)

of xt andxfdt

. xfdt

total frequency dependence of magnetic susceptibilityOne way to determine fdl and fdp is to solve explicitly for is the same as xfd

prexp and

xl : x

fdp

pedogenic frequency dependence of magnetic susceptibilityxfdl

lithogenic frequency dependence of magnetic susceptibilityxp= (xfdt

− fdlxt)/(fdp− fdl ) , (4) fdp pedogenic per cent frequency dependence

fdl lithogenic per cent frequency dependencexl= (xfdt

− fdpxt)/(fdl− fdp) . (5) x

b collective theoretical lithogenic susceptibility

Fed iron extractable in CBDThe quantities fdl and fdp are constants and can be estimated Fet total ironfrom the slopes of regression lines between pre- and post-CBD

© 2000 RAS, GJI 142, 551–562


Figure 1. Variations in magnetic properties and iron fractions for parts of the Malan/Lishi formations at Jiaodao. (a) Total low-frequency magneticsusceptibility; (b) CBD-extractable and post-CBD low-frequency magnetic susceptibility; (c) pre-CBD per cent frequency-dependent magnetic

susceptibility; (d) free secondary iron (CBD-extractable); (e) total iron; and (f ) ratios of free secondary iron to total iron.

as shown by the Fed/Fet ratio (Fig. 1), a greater percentage of CBD affects natural lithogenic magnetite. We found that the

largest reduction in susceptibility occurred after the first steptotal iron is in the form of free secondary iron in palaeosols(40–60 per cent) than in loess (<40 per cent). (Fig. 2), which is consistent with the results of Hunt et al. (1995)

for synthetic samples. After that, the effect on susceptibilitydecreased significantly and was progressively smaller with eachsubsequent step. These results suggest that there is almost

3.1.2 T he eVects of CBD extraction on natural lithogeniccomplete removal of pedogenic magnetic carriers after the first

magnetitestep, and complete removal after the second step of thetreatment. The decreasing susceptibility reductions after sub-Using stepwise CBD treatment, Hunt et al. (1995) showed that

the standard CBD procedure completely dissolves synthetic sequent steps are ascribed to slow dissolution of fine-grainedlithogenic magnetite. Based on these results, it appears that,magnetite grains smaller than 0.2 mm and that it removes

40–60 per cent of the magnetic susceptibility from synthetic in the future, when the CBD treatment is used to separate

pedogenic and lithogenic components in a loess/palaeosol1 mm magnetite grains. We have used the same stepwiseapproach on natural samples from Jiaodao and from North sample, one or two additions of dithionite might be sufficient

(1 or 2 g of dithionite for 4 g of sample).American loess. Our goal was to evaluate the degree to which

© 2000 RAS, GJI 142, 551–562


Figure 2. Results of stepwise CBD extraction of several North American loess samples and one sample from palaeosol S5 (depth 46 m). Initiallow-frequency magnetic susceptibility was measured before the extraction. Other bars show remaining magnetic susceptibility after each consecutive

extraction step. The first and the second steps of the treatment appear to remove the pedogenic component; the reduction after other steps can be

ascribed to the slow dissolution of fine-grained detrital magnetite.

lack of a correlation is that the CBD procedure has removed3.2 Mixing analysis

some lithogenic susceptibility carried by detrital grains at theThe mixing analysis requires that we have values for the per cent SP/SD (superparamagnetic/single-domain) boundary, where thefrequency dependence of the pedogenic component fdp and the frequency dependence is the greatest. However, the explanationper cent frequency dependence of the lithogenic component may also be related to the instrumentation. The post-CBDfdl . The method will only work if there is a significant difference susceptibilities are only one order of magnitude larger thanbetween the two parameters. the detection limits of the Bartington bridge. Even though weTo obtain an estimate of fdp , we add and subtract fdp

xl to made four consecutive measurements of both low- and high-

eq. (3) and rewrite it as frequency susceptibility and four consecutive blank measure-

ments in between, our data set still included many negativexfdt

= fdpxt− (fdp−fdl )

xl . (6)

values for xfdpost

. Such values are physically impossible andAs noted above, xpost provides a reasonable approximation of can only be explained by noise and instrumental drift. Whenxl . Furthermore, for the Jiaodao sequence,

xpost is constant to we set the negative values to 0, the correlation between xpostfirst order so that the second term on the right-hand side of

and xfdpost

improved (Fig. 3b). Even though the regression lineeq. (6) can be taken as a constant. In that case, the slope of the

is not statistically significant because of the high scatter in theregression line between the pre-CBD magnetic susceptibility

data, there is a positive correlation with the slope of 2 per(xt=

xpre ) and its frequency dependence (

xfdt

=xfdpre

) can becent. This value is close to that obtained from the pre-CBD

used to determine fdp . Forster et al. (1994) defined this slope frequency dependence of the least weathered loess samples,as Fc , the collective true frequency dependence, and found which can be regarded as an independent estimate. We there-that it varies between 9.4 per cent and 12.4 per cent for

fore used 2 per cent as the value for fdl in the mixing analysis.soil sequences in the Chinese loess plateau, Tajikistan andWe recognize that this value is not well-constrained andCalifornia. The highest value was that of the Luochuan section.probably represents only an upper limit for fdl .For Jiaodao, the regression analysis gives a value of 12.1 perFor the palaeosols, the values of xp calculated from thecent (Fig. 3a). Furthermore, all samples in the data set have a

mixing analysis are generally in very good agreement withper cent frequency dependence lower than this value, which isthe values of xext determined by the CBD procedure (Fig. 4).consistent with the assumption that fdp is the per cent frequency Where there are differences, xp is mostly lower, but usually bydependence of purely pedogenic material.only a few per cent and only rarely by more than 20 per cent.Whereas it is relatively simple to determine fdp , it is more In contrast, for the loess, the values of xp are often somewhatdifficult to estimate fdl , the per cent frequency dependence lower than the corresponding values of xext , especially forof the lithogenic component. One possible approach wouldthe upper two loess layers (L1 and L2). There xp generallybe to use the relationship xfd

l

= fdlxl and the fact that

xl and

represents only 40–80 per cent of xext . These differencesxfdl

can be approximated by xpost andxfdpost

, respectively.between the two sets of estimates can be explained by theHowever, we found no statistically significant relationship

between xpost andxfdpost

. One possible explanation for the removal of lithogenic magnetite by the CBD procedure.

© 2000 RAS, GJI 142, 551–562


able, and the relative peak heights conform to published data,

especially those reported for the Luochuan section (Fig. 1)(Kukla et al. 1988; Kukla & An 1989).The amounts of total (Fet) and pedogenic iron (Fed ) are

higher in palaeosols than in loess (Fig. 1). One could arguethat the susceptibility enhancement in palaeosols could beascribed to the higher total iron content. However, the ratio

of total to pedogenic iron shows that a higher percentage oftotal iron has been transformed to pedogenic Fe-compoundsin palaeosols than in loess. This confirms that the higher

magnetic susceptibility in the palaeosols can be ascribed largelyto in situ pedogenic formation of ferrimagnetic minerals andnot to variations in the amounts of lithogenic ferrimagnetic

minerals.Although the estimates of lithogenic susceptibility based on

the CBD procedure are generally lower and less variable than

those based on the mixing analysis, both data sets show along-term trend to lower values as loess and palaeosols becomeolder (Fig. 5). The long-term decline in estimated lithogenic

susceptibility suggests that palaeoclimatic interpretations basedonly on total susceptibility will be in error and that a singleestimate of average lithogenic susceptibility is not an accurate

basis for adjustment of the total susceptibility. This long-termtrend has several possible explanations: (1) a systematic but

gradual change in the lithogenic magnetic mineralogy of loessfrom older to younger units; (2) a gradual increase in depositionof a minor CBD-resistant component such as cosmic dust or

volcanic ash; or (3) in situ weathering of lithogenic magnetite,perhaps not only during pedogenesis but also after burial. Wefavour the hypothesis of in situ oxidation of lithogenic mag-netite, either by more intense or longer periods of weatheringin the lower portions of the sequence, or by a combination ofsurface weathering followed by slow long-term post-burial

oxidation. The idea of more intense or longer periods ofweathering in the lower portions of the sequence is consistentFigure 3. Determination of mixing analysis parameters. (a) Theoreticalwith the presence of the S5 palaeosol at the bottom of thepedogenic per cent frequency dependence (fdp) is determined from thestudied section. This unit is generally believed to representslope of the regression line between the pre-CBD frequency dependence

of magnetic susceptibility (xfdpre

) and the pre-CBD low-field mag- the climatic optimum for the past 1.2 Myr (An et al. 1987),netic susceptibility (xpre). (b) Theoretical lithogenic per cent frequency although lately its advanced degree of weathering has beendependence (fdl ) is estimated from the slope of the regression line ascribed to prolonged periods of weathering rather than tobetween the post-CBD frequency dependence of magnetic susceptibility more intense ones (Han et al. 1998). The hypothesis of in situ(xfdpost

) and the post-CBD low-field magnetic susceptibility (xpost ). oxidation of lithogenic magnetite can be tested by determining

the lithogenic susceptibility farther down in the loess/palaeosolsequence.

The values of xl calculated from the mixing analysis arehigher and show greater variability than values of xpost obtained 4.2 Comparison of CBD procedure and mixing analysisfrom the CBD procedure (Fig. 5). A weighted least-squaresprocedure with a 10 per cent smoothing population was used Estimates of the pedogenic susceptibility determined from

CBD data and from the mixing analysis are very similar. Into reduce the variability. The smoothed curves for xl andxpost show a similar long-term decrease with depth in the fact, the cross-correlation coefficient between xext and

xp is

0.997 (Fig. 6a). Furthermore, the pedogenic component, assection but there are higher-frequency variations in the xl curvethan in the xpost curve. determined by either method, is the primary factor in deter-

mining the shape of the susceptibility curve, which confirms

that the susceptibility enhancement is indeed of pedogenic4 DISCUSSION

origin.In the palaeosols, the pedogenic susceptibility peaks from

4.1 Implications for interpretation of the palaeoclimaticthe mixing analysis (xp) tend to be slightly lower than thoserecordfrom the CBD procedure (xext). Two factors are involved here.The first is that xext may be an overestimate of the actualThe total low-field magnetic susceptibility curve for the

Malan/Lishi part of the Jiaodao section is similar to curves pedogenic susceptibility because of the dissolution of lithogenicmagnetite by the CBD procedure. The second is that xp maypublished for other loess/palaeosol sequences (Fig. 1). Each of

the major palaeosol and loess couplets is clearly distinguish- be an underestimate of the actual pedogenic susceptibility

© 2000 RAS, GJI 142, 551–562


Figure 4. Pedogenic magnetic susceptibility as estimated from (a) the CBD procedure (xext ) and (c) mixing analysis (xp). (b) Comparison of CBD

(black line) and mixing analysis (grey line) estimates.

because the value we used for fdl is probably an upper limit. 4.2.1 Adjustment of the CBD estimatesNevertheless, because the lithogenic component makes only asmall contribution to the total susceptibility of the palaeosols, As noted above, the CBD procedure dissolves some litho-

genic magnetite and this leads to overestimation of thethe peaks are not significantly changed if a correction is madefor the effects of CBD on lithogenic magnetite or if there is a pedogenic component. Although the amount of extracted litho-

genic susceptibility varies from sample to sample, we cansmall adjustment in the parameters of the mixing model.In the loess layers, the lithogenic susceptibility removed by estimate an average amount of removed lithogenic susceptibility

by comparing the collective average lithogenic susceptibilitythe CBD procedure or the uncertainties in the lithogenic

susceptibility arising from uncertainties in fdl represents a (xb of Forster et al. 1994) derived from the intercept of theregression between the pre-CBD susceptibility and its fre-non-negligible contribution to the total apparent pedogenic

susceptibility. However, even when these factors are taken into quency dependence (20.5×10−8 m3 kg−1) (Fig. 3a) with thepost-CBD susceptibility (xpost ) averaged over the entire sectionaccount (see below), the pedogenic susceptibility component

of the loess layers is still considerable (Fig. 4). These results (17.2×10−8 m3 kg−1 ). As expected, the latter is lower becauseof the removal of lithogenic magnetic susceptibility.confirm our earlier findings (Verosub et al. 1993) that the

susceptibility in the ‘unweathered’ loess is both lithogenic and To compensate for the effects of the CBD procedure onlithogenic magnetite, we multiplied xpost by the CBD correctionpedogenic, and that pedogenesis occurs even during periods

dominated by loess deposition. factor of 1.2 (20.5/17.2). Although this factor does not account

© 2000 RAS, GJI 142, 551–562


Figure 5. Lithogenic magnetic susceptibility as estimated from the CBD procedure (xpost ) and mixing analysis (xl ). (a) Lithogenic component

determined by CBD (xpost , black line) and adjustment for removal of some of the lithogenic magnetite (Adj.xpost , grey line) (see Section 4.2.1).

(b) Lithogenic component determined by mixing analysis (xl , black line) and adjustment for overestimation of fdl parameters (Adj.xl , grey line)

(see Section 4.2.2). Both sets of data are fitted by a weighted least-squares curve with a 10 per cent smoothing population. (c) Comparison of

estimates from CBD (xpost , black line) and mixing analysis (xl , grey line). (d) Comparison of adjusted CBD estimates (Adj.

xpost , black line) and

adjusted mixing analysis estimates (Adj.xl , grey line).

for differences in the amount of fine-grained magnetite in dence of the lithogenic component. We then adjusted fdl to1 per cent. That decreased the mixing analysis estimates byindividual samples, it reduces the magnitude of the effect of

the dissolution of lithogenic magnetite and brings the estimates 10 per cent of their previous value and brought them closer tothe corrected CBD values. The average of adjusted xl valuesbased on the CBD procedure closer to the mixing model

estimates. from the mixing analysis decreased to 21.7×10−8 m3 kg1.The adjustment lowered the estimates of the lithogenic

component but did not reduce the variability. This suggeststhat the mixing model may be very sensitive to slight differences

4.2.2 Adjustment of the mixing analysis estimatesbetween the true per cent frequency dependence and the valueused in the mixing model. For example, if the pedogenicWe also compared the average value of the lithogenic

susceptibility from the mixing analysis (23.7×10−8 m3 kg−1 ) conditions that affected the least altered loess strata and the

most highly altered palaeosols deviated significantly fromwith the value xb (20.5×10−8 m3 kg−1) from the regressionanalysis (Fig. 3a). The result is consistent with the hypothesis the mean, a systematic error could lead to clusters of outliers

(TenPas 1997). In addition, the xl estimates may be verythat we have overestimated fdl , the per cent frequency depen-

© 2000 RAS, GJI 142, 551–562


Then the average of the calculated xl values can be comparedwith xb , and fdl can be adjusted accordingly. Iteration cancontinue until the average of xl reaches a value close to

xb .

4.2.3 Comparison of adjusted estimates

The adjustment of both sets of pedogenic susceptibility

estimates improves their agreement significantly. If both sets

of estimates were exactly the same, they would plot on a

regression line with intercept 0, slope 1, and correlation

coefficient 1. The original estimates correlate well (slope

0.99, r=0.997) (Fig. 6a), but the intercept is negative(−5.96×10−8 m3 kg−1 ). We have shown in Sections 4.2.1 and4.2.2 that the CBD procedure results in an overestimation of

the pedogenic component while the mixing model estimates

results in an underestimation. The adjustment of both sets of

data does not change the already high correlation coefficient

and slope, but it brings the intercept very close to the origin

(−0.11×10−8 m3 kg−1 ), suggesting that the values are verysimilar and that they both represent good estimates of the

pedogenic component (Fig. 6b).

4.3 Magnetic mineralogy of the pedogenic component

In Fig. 7, we compare our two estimates of the pedogenic

susceptibility with the CBD-extractable iron (Fed). We findthat we can divide the secondary iron into three components:

(1) a part that is correlated with the estimate for the pedogenic

susceptibility; (2) a part that does not carry a magnetic

susceptibility; and (3) a part that is responsible for variations

in the other two components. The first component is clearly

the most important part in terms of interpreting a climate

signal because it is directly related to pedogenesis. The second

component is present in both loess and palaeosol samples andFigure 6. (a) Comparison of estimates of pedogenic magneticappears to represent secondary iron oxides that were presentsusceptibility from the CBD procedure (xext ) and mixing analysis (

xp). in the loess at the time of deposition on the loess plateau. ThisThe negative intercept suggests that the CBD estimates may be too

background secondary iron may originate from weathering inhigh or the mixing analysis estimates too low or both. (b) Comparison

of adjusted estimates of pedogenic magnetic susceptibility from the the source areas of the loess. The third component, whichCBD procedure (Adj.xext ) and mixing analysis (Adj.

xp) shows a good produces the scatter in Fig. 7, probably arises from natural

agreement between the two sets of estimates. variability in the process of pedogenesis. It may also reflect

the removal of lithogenic material by the CBD procedure.

The first component is produced by pedogenesis on the loess

plateau, and it carries the pedogenic susceptibility signal (Fig. 4).

To estimate the per cent of Fed that is needed to producesensitive to slight random errors in the measurements. Tothis susceptibility signal, we first observe that the pedogenicreduce these errors, the estimates of lithogenic susceptibilitysusceptibility increases by about 140×10−8 m3 kg−1 for acan be smoothed using a locally weighted least-squares curve1 per cent increase in Fed (Fig. 7). We then calculate theapproach with a 10 per cent smoothing population as showncontribution to the magnetic susceptibility for a 1 per centin Fig. 5. The removal of random variability by smoothingincrease in Fed resulting from pure magnetite or pure maghemite.does not remove the high-frequency variation of the mixingThis contribution is equal to the inherent susceptibility of amodel lithogenic component estimates not shown by the CBDferrimagnetic material divided by the per cent of iron in theprocedure.material (70 per cent for maghemite and 72 per cent forThe adjustment of the mixing model component estimatesmagnetite). For larger grains of maghemite and magnetite,shows that the model can be employed without doing CBDthe inherent susceptibilities are 40 000×10−8 m3 kg−1 andextractions, as both the pedogenic proportionality factor fdp50 000×10−8 m3 kg−1, respectively. For superparamagneticand the lithogenic proportionality factor fdl can be estimatedgrains, these values should be doubled (Maher 1998) orfrom pre-CBD magnetic susceptibility measurements. Theincreased several-fold (Worm 1998).determination of the former is described in Section 3.2. TheWe can obtain a lower limit on the ferrimagnetic fractionlatter can be estimated as described here. First, the components

of Fed by assuming that all of the pedogenic material is super-are calculated using the frequency dependence per cent of theleast weathered loess sample as a first approximation of fdl . paramagnetic and that the inherent magnetic susceptibility of

© 2000 RAS, GJI 142, 551–562


Because the inherent susceptibilities of antiferromagnetic

minerals are of the order of 50–100×10−8 m3 kg−1 their con-tribution to the pedogenic susceptibility is negligible compared

with that of maghemite or magnetite.

4.4 Magnetic mineralogy of the lithogenic component

The concentrations of magnetic minerals carrying the litho-

genic magnetic susceptibility component can be estimated

from the magnetic susceptibility of magnetite, the magnitude

of lithogenic magnetic susceptibility, and the amount of CBD-

resistant Fe. The CBD-resistant Fe represents 60–70 per cent

of total Fe in loess, and 40–60 per cent in palaeosols (Fig. 1),

which is 1.3–2.5 per cent of the total mass of the sample. The

magnetic susceptibility of this fraction is 9–28×10−8 m3 kg−1(Fig. 5). This relatively low value indicates that most of

this iron must be paramagnetic and probably occurs as primary

iron silicates. Based on the concentrations of CBD-resistant

iron and on the magnetic susceptibility of paramagnetic

minerals (Thompson & Oldfield 1986), we estimate that the

maximum contribution of paramagnetic minerals cannot exceed

2×10−8 m3 kg−1. The remaining magnetic susceptibility mustcome from lithogenic ferrimagnetic minerals, which in this case

is relatively coarse-grained magnetite. Using the same pro-

cedure as in the previous section, we found that a 1 per cent

concentration of iron in the form of coarse-grained magnetite

would make a contribution of about 695×10−8 m3 kg−1 tothe magnetic susceptibility. Thus, lithogenic magnetite can

represent only about 2–3 per cent of the CBD-resistant Fe and

1–2 per cent of Fet .

Figure 7. Regressions between estimates of pedogenic magneticsusceptibility and free secondary iron (Fed). (a) Pedogenic component 5 CONCLUSIONSdetermined from the CBD procedure (xext). (b) Pedogenic component Our results confirm that the CBD procedure removes somedetermined by the mixing analysis (xp). Both regression lines suggest lithogenic susceptibility from both loess and palaeosol samplesthat magnetic susceptibility increases by about 140×10−8 m3 kg−1

and that it underestimates the lithogenic component andfor one per cent increase in secondary iron.

overestimates the pedogenic component. The magnitude of the

effect is moderately high in loess layers but almost negligible

in palaeosols. As a result, there is almost no effect on the peak

pedogenic susceptibilities, which are used for climatic inter-superparamagnetic grains is twice as high as that of larger grains.pretations. A correction factor can be used to adjust for theIn that case, the contribution to the magnetic susceptibilityunderestimation of the lithogenic component. By alteringwould be about 1140×10−8 m3 kg−1 for pure maghemite andthe CBD procedure somewhat, it might be possible to reduce1390×10−8 m3 kg−1 for pure magnetite, and about 10–12 perthe magnitude of the effect.cent of the Fed is needed to account for the pedogenic The mixing analysis also produces good estimates for thesusceptibility. However, if magnetic susceptibility of superpara-pedogenic component, especially for the peak palaeosol values.magnetic grains is several times that of larger grains (WormThe estimates of the lithogenic component based on the mixing1998), considerably less than 10 per cent of Fed could account analysis show considerably higher variability than those basedfor the pedogenic magnetic susceptibility.on the CBD procedure.For an upper limit, we assume that only half of theBoth methods show a long-term decline in lithogenic mag-pedogenic material is superparamagnetic. In that case, the

netic susceptibility with depth and hence with increasing agescontributions become 860×10−8 m3 kg−1 for maghemite andof the loess and palaeosols. This decline is probably due to1040×10−8 m3 kg−1 for magnetite, and about 13–16 per centmore intense or prolonged periods of weathering associatedof the Fed is needed to account for the pedogenic susceptibility.with the formation of older soils.Thus, it appears that only about 10–15 per cent of the FedThese variations in lithogenic susceptibility show that acorresponds to pedogenic maghemite or magnetite. The

single estimate of average lithogenic susceptibility should notremainder of the Fed must correspond to the other class ofbe used to adjust the total susceptibility curve, and that themagnetic minerals removed by the CBD procedure, namelyrelationship between climate and magnetic susceptibility mayantiferromagnetic minerals such as haematite and goethite.

These minerals also contribute to the pedogenic susceptibility. be more complex than has been generally assumed.

© 2000 RAS, GJI 142, 551–562


Fine, P., Singer, M.J., La Ven, R., Verosub, K. & Southard, R.J., 1989.Although the mixing analysis provides a sensitive, rapidRole of pedogenesis in distribution of magnetic susceptibility in twoand easily applied method for separating the lithogenic andCalifornia chronosequences, Geoderma, 44, 287–306.pedogenic components of the magnetic susceptibility in loess/Fine, P., Singer, M.J., Verosub, K.L. & TenPas, J., 1993. New evidence

palaeosol sections, the CBD procedure has the advantagefor the origin of ferrimagnetic minerals in loess from China, Soil

that it can be used to determine total Fe-partitioning between Sci. Soc. Am. J., 57, 1537–1542.lithogenic and pedogenic phases as well as to calculate the Fine, P., Verosub, K.L. & Singer, M.J., 1995. Pedogenic and lithogenicrelative contributions of individual magnetic minerals. In contributions to the magnetic susceptibility record of the Chinese

loess/paleosol sequence, Geophys. J. Int., 122, 97–107.general, the minerals responsible for the pedogenic magneticForster, T., Evans, M.E. & Heller, F., 1994. The frequency dependencesusceptibility signal represent less than 15 per cent of the CBD-of low-field susceptibility in loess sediments, Geophys. J. Int., 118,extractable iron; the remainder of this iron must be in the636–642.

form of antiferromagnetic minerals, which carry a negligibleHan, J.M., Lu, H.Y. & Guo, Z.T., 1996. The magnetic susceptibility of

susceptibility signal. The minerals responsible for the lithogenicmodern soils in China and its use for paleoclimatic reconstructions,

magnetic susceptibility represent less than 3 per cent of the Studia Geophys. Geodaetica, 40, 262–275.CBD-resistant iron; the remainder of this iron resides in Han, J.T., Fyfe, W.S. & Longstaffe, F.J., 1998. Climatic implications

paramagnetic silicate minerals. of the S5 paleosol complex on the southernmost Chinese loess

plateau, Quat. Res., 50, 21–33.A final advantage of the CBD procedure is that pre- andHeller, F. & Liu, T.S., 1986. Paleoclimatic and sedimentary historypost-CBD measurements of other environmental magneticfrom magnetic susceptibility of loess in China, Geophys. Res. L et.,parameters, such as ARM, IRM, SIRM, and hysteresis para-13, 1169–1172.

meters, can be used to obtain additional information aboutHeller, F., Shen, C.D., Beer, J., Liu, X.M., Liu, T.S., Bronger, A.,

the nature of the pedogenic and lithogenic magnetic minerals. Suter, M. & Bonani, G., 1993. Quantitative estimations of pedo-In the end, the purpose of a given study will probably be genic ferromagnetic formation in Chinese loess and paleoclimaticthe major factor in the choice of a separation method: if implications, Earth planet. Sci. L et., 114, 385–390.

Hunt, C.P., Singer, M.J., Kletetschka, G., TenPas, J. & Verosub, K.L.,palaeoclimatic interpretations are the main focus of the study,1995. Effect of citrate-bicarbonate dithionite treatment on fine-mixing analysis will provide accurate results in a very shortgrained magnetite and maghemite, Earth planet. Sci. L ett., 130,time without the need for CBD extractions. More detailed87–94.

studies of the concentration, magnetic mineralogy and mag-Hus, J.J. & Han, J., 1992. The contribution of loess magnetism in

netic properties of the individual components will require theChina to the retrieval of past global changes—some problems, Phys.

use of the more time-consuming CBD procedure. However, as Earth planet. Inter., 70, 154–168.we have shown, a combination of the two approaches probably Janitzky, P., 1986. Citrate-bicarbonate-dithionite (CBD) extractable

provides the most powerful and the most accurate way of iron and aluminum, in Field and L aboratory Procedures in a Soil

Chronosequence Study, eds Singer, M.J. & Janitzky, P., USGS Bull.,separating the magnetic susceptibility components.1684, 38–41.Kukla, G. & An, Z., 1989. Loess stratigraphy in Central China,

Palaeogeog. Palaeoclimat. Palaeoecol., 72, 203–225.Kukla, G., Heller, F., Ming, L.X., Chun, X.T., Sheng, L.T. &ACKNOWLEDGMENTSSheng, A.Z., 1988. Pleistocene climates in China dated by magnetic

Financial assistance was provided by NSF grants EAR-92–05191 susceptibility, Geology, 16, 811–814.Kukla, G., An, Z.S., Melice, J.L., Gavin, J. & Xiao, J.L., 1990. Magneticand EAR-97–10051 awarded to KLV and MJS. We thanksusceptibility record of Chinese loess, T rans. R. Soc. Edin.—Ea. Sci.,Yuan Baojin and Zhao Xiaolin, from the Chinese Academy of81, 263–288.Sciences in Beijing, who organized and led sample collection,Liu, T.S., ed., 1988. L oess in China, 2nd edn, Springer-Verlag, Berlin.

and Jeff Light who performed the determination of total Fe asLiu, X.M., Rolph, T., Bloemendal, J., Shaw, J. & Liu, T.S., 1995.

well as some magnetic susceptibility measurements and CBDQuantitative estimates of paleoprecipitation at Xifeng, in the loess

extractions. plateau of Central China, Palaeogeog. Palaeoclimat. Palaeoecol.,

113, 243–248.Maher, B.A., 1998. Magnetic properties of modern soils and

Quaternary loessic paleosols: paleoclimatic implications, Palaeogeog.REFERENCES Palaeoclimat. Palaeoecol., 137, 25–54.

Maher, B.A. & Thompson, R., 1991. Mineral magnetic record of theAn, Z.S., Liu, T., Zhou, Y. & Sun, F., 1987. The paleosol complex S5

Chinese loess and paleosols, Geology, 19, 3–6.in the China Loess Plateau—a record of climatic optimum during

Maher, B.A. & Thompson, R., 1992. Paleoclimatic significance of thethe last 1.2 Ma, Geojournal, 15, 141–143.

minerals magnetic record of the Chinese loess and paleosols, Quat.An, Z.S., Kukla, G.J., Porter, S.C. & Xiao, J., 1991. Magnetic Res., 39, 155–170.susceptibility evidence of monsoon variation on the Loess Plateau Maher, B.A. & Thompson, R., 1995. Paleorainfall reconstructions fromof central China during the last 130 000 years, Quat. Res., 36, 29–36. pedogenic magnetic susceptibility variations in the Chinese loessBanerjee, S.K., Hunt, C.P. & Liu, X.M., 1993. Separation of local and paleosoils, Quat. Res., 44, 383–391.signals from the regional paleomonsoon record of the Chinese loess Mehra, O.P. & Jackson, M.L., 1960. Iron oxide removal from soils andplateau: a rock magnetic approach, Geophys. Res. L et., 20, 843–846. clays by dithionite-citrate system buffered with sodium bicarbonate,Beer, J., Shen, C.D., Heller, F., Liu, T.S., Bonani, G., Dittrich, B. & Clays clay Min., 7, 317–327.Suter, M., 1993. 10Be and magnetic susceptibility in Chinese loess, Rutter, N., 1992. Presidential address, XIII INQUA Congress: ChineseGeophys. Res. L et., 20, 57–60. loess and global change, Quat. Sci. Rev., 11, 275–281.Evans, M.E. & Heller, F., 1994. Magnetic enhancement and paleo- Singer, M.J., Bowen, L.H., Verosub, K.L., Fine, P. & TenPas, J., 1995.climate: study of a loess/paleosol couplet across the loess plateau of Mossbauer spectroscopic evidence for citrate-bicarbonate-dithionite

extraction of maghemite from soils, Clays clay Min., 43, 1–7.China, Geophys. J. Int., 117, 257–264.

© 2000 RAS, GJI 142, 551–562


TenPas, J.D., 1997. Revised methods for paleoclimatic interpretation Verosub, K.L., Fine, P., Singer, M.J. & TenPas, J., 1993. Pedogenesis

and paleoclimate: Interpretation of the magnetic susceptibility recordof magnetic susceptibility in the Chinese loess sequence,MSc thesis,

University of California, Davis. of Chinese loess-paleosol sequences, Geology, 21, 1011–1014.Worm, H.U., 1998. On the superaparamagnetic-stable single domainTenPas, J.D., Vidic, N.J., Singer, M.J. & Verosub, K.L., 1999. Mineral

magnetic and pedogenic studies of the paleoclimate record of the transition for magnetite, and frequency dependence of susceptibility,

Geophys. J. Int., 133, 201–206.upper part of the loess/paleosol Sequence at Jiaodao, Chin. Sci.

Bull., 44, Suppl. 1, 259–263. Zhou, L.P., Oldfield, F., Wintle, A.G., Robinson, S.G. & Wangli, J.T.,

1990. Partly pedogenic origin of magnetic variations in ChineseThompson, R. & Oldfield, F., 1986. Environmental Magnetism, Allen

& Unwin, London. loess, Nature, 346, 737–739.

© 2000 RAS, GJI 142, 551–562

Separation of pedogenic and lithogenic components of...

Documents

Transcript of Separation of pedogenic and lithogenic components of...