中国科技论文在线 http://www.paper.edu.cn

Science in China Series D: Earth Sciences www.scichina.com

© 2008 SCIENCE IN CHINA PRESS earth.scichina.com www.springerlink.com Springer

Two pedogenic models for paleoclimatic records of from Chinese and Siberian loess

LIU XiuMing1,2†, LIU TungSheng3, Paul HESSE2, XIA DunSheng4,1, Jiri CHLACHULA4 & WANG Guan1

1 Key Laboratory of Western China’s Environmental Systems, Ministry of Education, Lanzhou University, Lanzhou 730000, China; 2 Department of Physical Geography, Macquarie University, NSW 2109, Australia; 3 Institute of Geology and Geophysics, Chinese Academy of Sciences, Beijing 100029, China; 4 Key Laboratory of Desert and Desertification, Institute of Cold and Arid Regions Environmental and Engineering, Chinese Academy of Sciences, Lanzhou 730000, China; 5 Laboratory for palaeoecology, Technical University Brno, FT VUT Zlin, 762 72 Zlin, Czech Republic The magnetic susceptibility (MS) of Chinese loess showing a general proportional relationship to pe- dogenic grade has been widely recognized and used for reconstruction of paleoclimate by Quaternary scientists. The in-situ pedogenic enhancement of ferrimagnetic content is normally believed to be the main reason for the increase of susceptibility in soil units. However, this pattern of high magnetic sus- ceptibility in palaeosols, and low values in loess, are not replicated in some loess deposits. Siberian loess deposits display a completely opposite susceptibility behavior: high values in loess and low values in palaeosols. This inverse relationship has been explained by the idea that magnetic suscepti- bility is reflecting the magnitude of an aeolian ferrimagnetic component of consistent mineralogy, the grain size of which is related to average wind velocity. Our magnetic study of Siberian samples in this paper suggests that there are notable differences in magnetic properties between Siberian loess and developed palaeosols, not only in magnetic grain-size and concentration but also in magnetic miner- alogy. This evidence is difficult to explain fully through variation in wind strength alone, but implies that the low magnetic susceptibility values in the Siberian paleosol units are a reflection, at least in part, of the alteration of the ferrimagnetic content by post-depositional processes. The Loess Plateau is a very arid area where potential evaporation is always higher than precipitation; pedogenesis occurs under dry oxidising conditions. The Siberian Kurtak region is located on the edge of the tundra where it is always wet and saturation during interglacials will lead to a reducing pedogenic environment. Ferri- magnetic minerals under this condition will be destroyed, resulting in lower magnetic susceptibility. Therefore, great care should be taken when using susceptibility values for paleoclimatic reconstruc- tion.

Siberian loess, Chinese loess, paleoclimate, magnetic susceptibility, , environmental

1 Introduction decades, not only have the loess deposits been dated but it was also found that magnetic susceptibility (MS, or χ) China has the thickest and most extensive loess deposits of loess and paleosol sequences shows a general propor- in the world. The Chinese loess consists of a number of tional relationship to pedogenic development[1] which loess layers deposited during dry-cold glacial periods, has a direct linkage to paleoclimate. This success of us- and interbeded soils developed during warm-moist in- Received January 23, 2007; accepted July 19, 2007 [1] terglacial periods . It is these loess and paleosol layers doi: 10.1007/s11430-007-0145-2 †Corresponding author (email: [email protected]) which provide a unique geological record for Quaternary Supported by the National Natural Science Foundation of China (Grant Nos. research of paleoclimatic change. During the last two 40421101 and 40772109) and Macquarie University, Australia

Sci China Ser D-Earth Sci | Feb. 2008 | vol. 51 | no. 2 | 284-293 转载 中国科技论文在线 http://www.paper.edu.cn

ing magnetic susceptibility as a paleoclimatic proxy in- values[22]. At a given site, the alternation between strong dicator from loess-paleosol sequences led to a revolution winds during glacial periods and reduced wind intensity in paleoclimatic studies from qualitative pedogenic de- during interglacial will therefore produce a fluctuating χ scription to quantitative or semi-quantitative reconstruc- signal. This hypothesis was subsequently used by tion. Magnetic susceptibility has also been used for Chlachula et al. to explain the similar susceptibility rela- correlation between continental and marine sedi- tionship with pedogenesis in Siberian loess[16,17]. Some [2,3] ments , for quantitative reconstruction of paleoclimate later researchers have reported some degree of pedogene- [4―7] (precipitation) , and even for reconstruction of pa- sis, however, they still in general accepted the above ‘wind leoprecipitation distribution for various time periods and velocity’ theory[17―20], because pedogenic features of the [8] spaces . Chinese Loess Plateau were not confirmed in Siberia. Similar cases of proportional pedogenic development Several questions arise concerning the interpretation of and MS values are also observed by other loess studies, these various susceptibility records. How do these curves - for example from the Czech Republic, Austria, Hungry, indicate paleoclimatic change? Is the magnetic susceptibil- [9] Poland, France and Germany . Several hypotheses have ity of Siberian loess determined by wind velocity? How to been put forward to interpret the relationship between explain the phenomenon of soil development and soil the MS of Chinese loess and paleoclimate, such as uni- gleying in well-developed Siberian soil[16,17]? Research on versal magnetic dust[10], and compression and leaching [11] magnetic properties of Alaskan loess by Liu et al. has re- during pedogenesis , but the more widely accepted ported that different magnetic minerals are present in the interpretation is in-situ formation of ultra-fine [23,24] loess and soil layers . They argued that the magnetic maghemite and by organic (bacterial origin) [12―15] signals recorded by Alaskan loess and paleosols cannot be and non-organic processes during pedogenesis to interpreted by the ‘wind velocity’ theory alone. Their ex- enhance soil MS. Aridity is a general feature in the Chi- planation emphasises the importance of pedogenesis[23,24], nese Loess Plateau: where potential evaporation is al- as an alternative hypothesis to explain the two different ways more than precipitation. This results in dry and behaviours but in one model. Their pedogenic-modifica- well oxidized topsoil conditions. During warmer and tion model acknowledges a difference of magnetic input moister interglacial periods, conditions would be suit- between glacial and interglacial but in addition that the able for fine iron oxide minerals, such as maghemite and lower susceptibility of paleosols in Alaska and Siberia was magnetite, to form in the topsoil leading to higher χ and caused by pedogenic destruction of ferrimagnetic minerals frequency dependent susceptibility (χfd). under reducing soil conditions. This paper reports on a However, magnetic susceptibility of Siberian loess detailed investigation of magnetic properties on selected profiles demonstrates an opposite magneto-clima- samples from Siberia to examine variations of magnetic tological signal: a negative correlation with pedogenic mineralogy from loess to welldeveloped soil. development (Figure 1). Magnetic maxima occur in gla- cial loess layers alternating with minima in interglacial 2 Samples and methods paleosols [16―20], quite similar to the behaviour reported for Alaskan loess[21]. The negative behaviour of susceptibility, The samples used for this paper comes from Kurtak, however, is actually very like that observed in marine Siberia[16], located in the valley of the Yenisey River in cores (low χ in interglacial), which are commonly inter- southern Siberia (55.1°N, 91.4°E). Total thickness of the preted as reflecting the variation in aeolian transport of loess is 32 m, its age covers Stages 1 to 7 of the marine dust to marine sites. This rationale has been extended to oxygen isotope[16,17] (Figure 1). The Kurtak section is the interpretation of the Alaskan loess[22]. Subsequent ex- typical of the area and many studies have previously periments led to the suggestion that susceptibility varia- been reported[16―20]. The age of 30 ka has been obtained tions reflected density sorting during aeolian transport. at a depth of 8 m (Stage 3) by 14C method[17] (Figure 1). Similar processes operating on a large scale during natural A total of 38 ages within Stage 3 at the Kurtak area have windstorms were suggested so that loess deposited near been reported, between 24 and 59 ka[17]. Some more the source region or by strong winds will be characterised dating was also reported for Stages 2 and 4 (see ref. [17] by higher χ values and coarser grains, whereas distal loess for more details). This section therefore has good age will be relatively depleted in magnetite and show lower χ control.

LIU XiuMing et al. Sci China Ser D-Earth Sci | Feb. 2008 | vol. 51 | no. 2 | 284-293 285 中国科技论文在线 http://www.paper.edu.cn

Figure 1 Correlation of loess/paleosol strata and magnetic susceptibilities from Kurtak, Siberia[17] and Xifeng, China[14]. The numbers on the right side are marine oxygen isotope stages; the description of left and bottom scales is given in ref. [17]. The samples collected for this study are shown as a dashed area, between stages 4―6.

This study will examine variations of mineralogy for hysteresis loop) and high temperature dependence of during pedogenesis of loess to soil, in particular to ob- saturation magnetisation (Ms(t)) were carried out in a serve if there is a phase change related to magnetic min- variable field transition balance (VFTB) in CSIRO Rock eralogy. Eight samples were collected from 15 to 22 m Magnetic Laboratory, Sydney. The VFTB measured Ms(t) depth for this study, with different susceptibility values under a field of 760 mT with heating rate 30℃/min and covering a range from loess to well-developed soils, a closed measuring environment of nitrogen gas to pre- (Figure 1, Stages 4―6, dashed line area). MS (χ) and vent oxidizing during heating. Low temperature mag- frequency-dependent susceptibility (χ ) were measured fd netisation measurement was carried out with a Molspin at Macquarie University, Sydney, using a Bartington MS Minispin at Macquarie University. Samples were first system. Isothermal remanent magnetization (IRM) ac- ℃ quisition was measured using a Molspin Minispin mag- cooled in liquid nitrogen to −196 then a field of 1.3 T netometer; magnetic field was generated by an ASC was applied, and then measured as they recovered to [6] Pulse magnetizer. Vibrating sample magnetisation (VSM room temperature (see Liu et al. ).

286 LIU XiuMing et al. Sci China Ser D-Earth Sci | Feb. 2008 | vol. 51 | no. 2 | 284-293 中国科技论文在线 http://www.paper.edu.cn

3 Rock magnetic measurements saturated traces even under an applied field of 2.3T, likely reflecting that every sample also contains hard 3.1 Magnetic properties at room temperature magnetic minerals, such as goethite, [20] or other As reported by previous studies, χ at Kurtak is high in weak magnetic and paramagnetic minerals such as iron [16] loess layers but low in paleosols (Figure 1). The χfd sulfides (see following text). The coercivity of rema- (Table 1) varies very little, so that its relation with soils nence (Hcr) is the reverse field required to reduce the cannot be clearly defined[16―20]. That may imply that remanent magnetisation to zero from saturation. It is a Kurtak magnetic minerals experienced different proc- useful parameter to provide information on the magnetic esses during pedogenesis from that of Chinese loess. softness/hardness and grain size. The remanence coer- civity (H ) of these 8 samples is between 46 and 65 mT Low values of χfd indicate absence or lower concentra- cr tion of extra fine of high χ magnetic minerals. However, (Table 1, Figure 2(b)). A paleosol sample had the highest it doesn’t exclude the possibility of producing weak and Hcr, indicating less soft magnetic minerals and higher paramagnetic magnetic minerals (with low χ) during percentage of hard magnetic minerals than that of weak pedogenesis, such as goethite[20], greigite and pyrite[26―30], soils and loess samples. Sample SB5, is the most devel- [31,32] oped soil with the lowest χ among these 8 samples and and pyrrhotite . That is because their χ and χfd can be almost ignored in comparison with magnetite and the highest Hcr. The other samples, although their χ are maghemite if one of these is present in the same sample. quite different, display only small difference in Hcr, but Both loess and paleosol samples acquired 90% or Hcr for SB2 seems to be unusually low. These character- more of their saturation isothermal remanence (SIRM) istics of Kurtak samples are unlike the behavior of [23] [24] intensity under a field of 0.3T, implying a soft (easier Luochuan and Xifeng samples from the Chinese saturated) magnetic component, such as maghemite Loess Plateau: their Hcr is proportional (or in reversal) to and/or magnetite, is a major carrier of magnetic rema- their SIRM and χ, demonstrating a regular and gradual nence. Low SIRM in soil while high in loess, suggests a change in grain size and soft/hard components. In com- higher concentration of soft magnetic components in parison, the data of Siberian samples do not show such a loess than in soils. Samples do not demonstrate fully systematic change. The ‘softest’ sample is found neither Table 1 Magnetic parameters for Siberian loess and paleosols χ Sample Loess Ms Mrs Hc χfd ΔMs SP SD MD SIRM Hcr −2 2 −2 2 −1 name /soil (10 Am /kg) (as left) (mT) (10−7 m3/kg−1) (%) (%) (%) (%) (%) (10 Am kg ) (mT) SB5 soil 2.05 0.37 12.8 11.58 1.77 −89.6 18.2 49.5 32.3 1.22 65 SB6 w. s 3.35 0.52 12.2 13.81 1.05 0 14.1 45.9 40 1.37 52 SB3 w. s 16.75 2.92 16 48 36 1.87 54 SB1 w. s 0.12 0.41 4.86 18.44 1.82 25.7 11 44 45 1.71 55 SB2 w. s 22.03 1.86 11.6 44.4 44 1.86 46 SB8 loess 22.36 0.19 11.3 37.7 51 1.78 52 SB7 loess 5.19 0.62 9.35 27.21 0.52 36.6 11.6 39.4 49 2.21 54 SB4 loess 6.26 0.89 11.6 30.02 0.98 32.4 13 46 41 3.01 52

Ms, Saturation magnetisation; Mrs/Ms: ratio of saturation remanent magnetisation/Ms; Hc: saturation coercivity; χ: susceptibility; χfd: fre- quency-dependence of susceptibility; ΔMs: magnetisation difference at 20℃ between heating and cooling; SP: superparamagnetic; SD: single domain; MD: multidomain; Hcr: coercivity remanence; w. s: weak soil.

Figure 2 SIRM acquisition curves (a) and their reverse field demagnetised curve (b) in Siberia Kurtak section.

LIU XiuMing et al. Sci China Ser D-Earth Sci | Feb. 2008 | vol. 51 | no. 2 | 284-293 287 中国科技论文在线 http://www.paper.edu.cn

in the most-developed soil SB5, nor in the weakest Verwey transition around −150℃ indicating presence of weathered loess SB4, but in intermediate (or weak) de- coarser size MD magnetite. The results indicate that veloped soil SB2. This feature implies that, at least in loess sample contains higher percentage of MD magnet- Kurtak, variation of magnetic mineralogy is probably ite than soil (Figure 4). For example, loess SB7 loses not a gradual change, but that a discontinuous variation more intensity but soil SB5 loses less LT SIRM as tem- of magnetic mineralogy may exist between the loess and perature increases. From the LT curves, it can be calcu- soils. This is difficult to explain by the ‘wind velocity’ lated that soil SB5 comprises approximately 50% SD theory alone. magnetic grains, about 40% in loess (SB7), and the The vibrating-sample magnetometer (VSM) allows weakly developed soil SB6 is intermediate between rapid measurement of the hysteresis parameters of mag- them. However, at temperatures higher than the transi- netic minerals. Typical curves of the hysteresis for Sibe- tion observed at −150℃, all curves are relatively straight rian loess/paleosols (Figure 3) show all selected samples and nearly flat, indicating only a very small amount of to be almost saturated under 0.3 T field and the higher the SIRM is carried by SP grains in these samples. the χ, the higher the saturation magnetisation (Ms). Fur- thermore, the ratio of Ms to χ under field of 0.1T (low field susceptibility[14]) is higher in loess than in soils, implying a greater concentration of soft magnetic min- erals, or a higher ratio of soft/hard, in the Siberian loess than in soils. Above 0.3T, the hysteresis loops become straight and almost parallel, implying their high field [14] susceptibility (ratio of Ms/field) is very close, reflect- ing a high level of similarity in magnetic materials in these samples.

Figure 4 The thermal demagnetization curves for 3 selected samples from Kurtak, Siberia.

LT susceptibility curves display similar characteristics to LT thermal demagnetization, including an obvious Verwey transition at −150℃ for both loess and soils (Figure 5), suggesting coarser grains of MD magnet- ite/maghemite present in all samples. Their distinctive features mainly appear in their curves above −150℃ through their various inclinations. Within this tempera- ture range between −150℃ and room temperature (RT), every susceptibility curve always decreases gradually as temperature increases. The difference is that loess SB7 has a smaller inclination but soil SB5 has a larger incli- Figure 3 Magnetic hysteresis loops for Siberian loess and soils from nation. Such regular decrease is affected by the intrinsic Kurtak. control of paramagnetic material: paramagnetic suscep-

tibility is inversely proportional to temperature. This 3.2 Magnetic properties at low temperature implies that soil samples contain higher concentration of The magnetic remanence carried by different grain size paramagnetic material than loess. If pedogenesis devel- categories of superparamagnetic (SP), single domain ops further, the susceptibility of these samples will de- (SD, here including peudo SD-PSD as well) and crease further, and their LT susceptibility curve will multi-domain (MD), can be separated according to their change toward increasing their characteristic of para- LT curve features[6]. All 8 samples show an obvious magnetism, the Verwey transition will become less clear

288 LIU XiuMing et al. Sci China Ser D-Earth Sci | Feb. 2008 | vol. 51 | no. 2 | 284-293 中国科技论文在线 http://www.paper.edu.cn

and eventually disappear. Instead, a completely para- The results of these experiments clearly indicate that magnetic curve showing χ inverse with temperature[14,18] there are different magnetic mineral components in the will appear. loess and well developed soil samples which differ not only in grain size and percentage of mineral amounts but also in magnetic mineral types (or phases). According to the ‘wind velocity’ theory, wind strength controls the magnetic component of the deposits by gravity sorting. Such an explanation obviously ignores post-depositional alteration of the magnetic minerals. If χ variation in Si- berian loess is only interpreted by wind force alone, then it is very clear that differences of magnetic mineralogy between loess and well-developed soil samples observed above in Figure 6, become very hard to explain. If χ is taken as an indicator of pedogenic (or weathering) de- Figure 5 Low-temperature susceptibility curves for selected samples velopment, a general tendency will be clearly found: from Kurtak, Siberia. with increase of pedogenesis (from a→b→c→d in

Figure 6), χ decreases. The amount of maghemite (lost 3.3 Magnetic properties at high temperature during heating) decreases gradually, eventually disap- High temperature thermomagnetic J(t) curves (Figure 6) pearing completely (Figure 6(c)). With further pe- show all samples having a Curie temperature (Tc) of dogenesis, some weakly magnetic materials such as iron 580℃, indicating the presence of magnetite in all sam- sulfides, which are oxidized into magnetite during heat- ples, which is identical to the characteristics of above LT ing and result in an increase of magnetisation (Figure susceptibility and LT SIRM demagnetization. Further- 6(d)), are formed. This phenomenon is present not only more, according to their increase or decrease of mag- in this study, but also in previous studies of Kurtek and netisation between heating and cooling, three different Bachat sections Siberia[18―20], in which high temperature categories of samples can be divided as follow. (1) susceptibility measurements showed similar characteris- Weakly weathered loess - heating and cooling curves are tics to those reported here. irreversable, the magnetisation of cooling curve is lower Such phenomenon of thermal magnetic measurement than before heating (Figure 6(a), (b)) due to thermally intensity increase after cooling is often observed in unstable maghemite being converted to hematite result- studies of lake and marine sediments under relative re- ing in a loss of magnetisation[20,24,25]. (2) Well (strongly) ducing conditions. In order to interpret this observation, developed soils―the features of heating and cooling Geiss and Banerjee used natural minerals such as pyrite [29] curves are also irreversable, but the magnetisation of and pyrrhotite for testing. These weakly magnetic and cooling curve is even higher than before heating (Figure paramagnetic iron sulfides were proved to have irrever- 6 (d)), implying weakly magnetic materials, such as iron sable behavior during heating: their heating curves dis- sulfides or other paramagnetic minerals, which are con- play clear χ increase at temperature 250―300℃ and verted to strongly magnetic magnetite (Tc = 580℃) as a 500℃ as they are all converted to magnetite after cool- result of oxidation during heating process, leading to ing (Tc=580℃) and show an increase of magnetisa- great increase of χ. The cooled samples of these well tion[29]. These thermal magnetic properties are quite developed soils display a black color (magnetite forma- similar to those of reported Siberian soils[18―20] and tion), instead of the reddish color (hematite) observed Alaskan soils[23,24]. Under reducing environments, iron for the loess and weak soils. (3) Weakly developed soil- oxides such as maghemite and magnetite become unsta- heating and cooling curves are reversible (Figure 6(c)), ble and are gradually dissolved to form iron sulfides this type of sample contains neither thermally unstable such as greigite and pyrite [26―30]. For example, magnet- maghemite (which loses intensity during heating) nor ite is dissolved and greigite formed in marine sediments iron sulfides (which increases in intensity during heat- near New Zealand[26], while dissolution of iron oxides ing), but dominant magnetite. and formation of iron sulfides was reported from

LIU XiuMing et al. Sci China Ser D-Earth Sci | Feb. 2008 | vol. 51 | no. 2 | 284-293 289 中国科技论文在线 http://www.paper.edu.cn

Figure 6 Thermal magnetic curves of loess and paleosols for Kurtak, Siberia (χ: 10−7 m3kg−1).

Spain[27]. Iron sulfides pyrite and pyrrhotite have been mates respectively. It is therefore analogous to the ex- identified in Czech loess[31, 32]. planation applied to the Chinese loess: the general fea- tures of the original dust material are controlled domi- 4 Discussions nantly by wind velocity and provenance but once ae- olian dust is deposited pedogenesis starts. The local re- Magnetic measurements have shown obvious difference dox environment of the topsoil will affect the formation of magnetic mineralogy between loess and well devel- of iron-bearing minerals. In the arid Loess Plateau of oped soils in Siberia. Such a difference cannot be inter- China pedogenic processes under oxidising conditions preted by the ‘wind velocity’ theory. That is because generate ferrimagnetic minerals which increase its sus- variation of wind velocity between glacial and intergla- ceptibility. However in the Kurtak area of Siberia, where cial will cause different physical characteristics such as the topsoil is always wet, pedogenic process occur under mineral concentration and grain size, but should not re- reducing or low oxygen conditions, so that existing high sult in a notable difference in mineralogy as observed by χ ferromagnetic iron oxides are destroyed and converted this study. This phenomenon, was identified previ- - to iron sulfides and hydrates, resulting in an inverse re- ously[18 20], but the authors did not recognise its impor- lationship of magnetic susceptibility with pedogenic tance and instead regarded it as something related to development. organic materials. As early as 2001, Zhu’s research group identified and If the results of both thermal magnetisation (Figure 6 reported iron sulfides pyrite or[31]/and pyrrhotite[32] in this study) and thermal magnetic susceptibility[18,19] for loess and soils from Czech loess deposits. They not only Siberian loess are carefully analyzed and compared, a measured magnetic properties of the samples but also general magnetic behavior can be recognized. Less obtained direct evidence from XRD and SEM. All of weathered loess samples (such as Figure 6(a), (b)): irre- their high temperature susceptibility curves of loess and versibly lose intensity during heating. With pedogenic soils display two small peaks at around 250―300℃ and development to weakly developed soil (such as Figure ℃ 6(c)): the samples act reversibly and do not lose or in- 500 , indicating iron sulfides are converted irreversibly crease intensity during heating. Finally, with further pe- at these temperatures, but the peak around 500℃ dogenic development to well-developed soils (such as showed clear distinction between samples. Their χ in- Figure 6(d)) these samples increase intensity irreversi- tensities after cooling down are always obviously higher bly. than before heating. This succession from weak to strong pedogenesis is Many thermal magnetic susceptibility curves have actually following a variation of magnetic mineral char- been measured at Kurtak section by previous studies, of acteristics, responding to glacial and interglacial cli- which 9 measurements showed increased χ during heat-

290 LIU XiuMing et al. Sci China Ser D-Earth Sci | Feb. 2008 | vol. 51 | no. 2 | 284-293 中国科技论文在线 http://www.paper.edu.cn

ing (6 from ref. [18]; 1 from ref. [19]; 2 from ref. [20]). χ together with low χfd in loess, and high χ with high χfd Only one was a loess sample[20] while the other 8 sam- in paleosols, indicates the formation of SP/SD grains ples were soils. Their thermal magnetic susceptibility during pedogenesis[12,13]. In Siberia, low temperature [18,20] behaviors are very similar to those reported for Czech thermal demagnetization (Figure 4) and χfd imply [31,32] loess and soil containing iron sulfides . almost no SP magnetite/maghemite. Although the χ val- The Loess Plateau of China is a region characterized ues the Kurtak section are high in loess and low in soils, mainly by its aridity. According to meteorological statis- χfd displays almost no variation. This phenomenon tics of average precipitation of the loess area of Shangxi gradually changes away from the Siberian Tundra. For Province is about 520 mm per year, but the potential example at Bachat section (54.5°N, 87.1°E) about 150 evaporation is 1620 mm. The annual effective precipita- km southwestern of Kurtak, χ is still similar to Kurtak, tion (= precipitation-potential evaporation) is always showing high value in loess and low in soils. But χfd negative. The moisture deficit is even more critical fur- shows a clear tendency of high values in soils and low ther toward the western part of the Plateau. Under this [18] values in loess layers , implying that pedogenic de- the pedogenic environment is oxidising and the velopment under wet condition may distort or destroy amount of maghemite and magnetite forming in situ is [14] some features of magnetic properties. Further away in proportional to the amount of rainfall . Therefore in [34] [9] Tajikistan and the Ukraine the relationship between the Loess Plateau, the susceptibility shows a generally the χ and pedogenesis is very close to that of Chinese positive correlation to paleo-precipitation. loess. Apparently the cold and wet tundra climate leads The Kurtak area of Siberia is located on the edge of to high χ in loess and low χ in soils in both Siberia and the Siberian Tundra. The amount of precipitation is low Alaska. but evaporation is also extremely low, so that topsoil In fact, an earlier report of higher χ in topsoil of there is in a high moisture status. Even in summer, only Alaska (Figure 1) by Begét and Hawkins[21], has re- the topsoil is thawed and the soil below remains frozen. vealed information contradicting the ‘wind velocity’ Rainfall is prevented from infiltrating by this frozen theory. Higher χ in topsoil is likely related to localised layer enhancing the wetness of the ground surface. Ac- good soil drainage and oxidizing conditions. Moreover cording to pedogenic investigations by Chlachula[17], Vlag’s measurements of Halfway House and Gold Hill- there are quite clear differences between Kurtak loess steps sections near Fairbanks[35] have also proved to and soils, not only in color but also in soil types, and have high χ together with high χ in all measured top- magnetic susceptibility. It can be conjectured that during fd soils. This is quite similar to Xifeng loess[14] in China, cold glacial periods topsoil around Kurtak received less implying that conflicting phenomenon to ‘wind velocity’ moisture and was relatively drier, probably in a condi- theory are also present in Alaska. Furthermore, a study tion of relatively oxidising or weakly reducing. During by Rosenbaum et al. reported geochemical data which these stages maghemite and magnetite were preserved. showed a maximum in the concentration of chemi- When climate changed toward warmer and more humid cally-immobile titanium within paleosol horizons[36]. Ti interglacial conditions the pedogenic environment of the concentration, an excellent proxy for detrital heavy topsoil became reducing. Beginning with smaller grains, mineral content, was consistent with leaching of mobile maghemite is slowly reduced to iron sulfides and hy- elements, principally Si. This result is also difficult to droxides (Figure 6) in accordance with its environment. explain by the variation of wind force alone. This eventually results in a negative relationship of These findings therefore lead to the conclusion that magnetic susceptibility with pedogenic development. different behaviours of magnetic susceptibility between According to the ‘wind velocity’ hypothesis, wind Chinese and Siberian loess are mainly produced by their strength affects the χ values of loess through physical different pedogenic environments and climatic condi- sorting of magnetic grains. This theory does not consider tions. In the central area of the Loess Plateau, such as at post-depositional alteration. Frequency-dependent sus- Luochuan and Xifeng, continual moisture deficit creates ceptibility χfd is regarded as a sensitive indicator of grain oxidising pedogenic conditions. The 5th paleosol layer sizes around 0.03 μm (around the boundary of SP and (S5) is the most developed soil on the whole section, SD for magnetite). In the Chinese Loess Plateau, low while the 1st paleosol layer (S1) is the second. Based on

LIU XiuMing et al. Sci China Ser D-Earth Sci | Feb. 2008 | vol. 51 | no. 2 | 284-293 291 中国科技论文在线 http://www.paper.edu.cn

their appearance of dark reddish brown colour and susceptibility curve will bear no relation with paleocli- analysis of magnetic mineralogy[1―14], their pedogenesis mate or pedogenesis. Such a state has been reported occurred under strongly oxidising conditions. The mag- from New Zealand and Argentina, and from the edge of [23,24,37] netic susceptibility for S5 and S1 gain the highest and the Chinese Loess Plateau . second highest values in their section respectively. Their high χ together with high χfd indicate many extra-fine 5 Conclusions grains of maghemite and magnetite, formed during pe- This study indicates that the magnetic susceptibility dogenesis under warm, moist paleoclimates. In the Kur- from Kurtak of Siberia was clearly affected by pe- tak area of Siberia however, pedogenesis results in de- dogenesis. Weakly weathered loess is representative of creasing χ in soils (and lower concentration of ferri- the original aeolian component whose magnetic compo- magnetic minerals) and increasing paramagnetic miner- sition is determined by provenance and by wind strength. [20] [31,32] als such as goethite and/or iron sulfides because Magnetite and maghemite are the two main contributors of the reducing conditions within interglacial soil pro- to the magnetic susceptibility. As climate varies toward files. These newly formed paramagnetic minerals have higher temperature and higher moisture, local pedogenic weak magnetism and their contribution to soil χ and χfd or weathering grade also increases. However pedogenic is therefore minimal. development under these conditions tends toward higher Based on the analysis above, the magnetic suscepti- reduction potential and, as a result, the iron oxides (such bility is simply a physical parameter showing no essen- as maghemite and magnetite) become unstable and are tial relationship with climate. Its value is controlled by gradually converted to iron hydrate and sulfides. This magnetic mineral type, concentration and grain size. The process decreases magnetic susceptibility, while weakly magnetic properties of detrital minerals are affected magnetic iron hydrates and sulfides are formed. greatly by post-depositional redox condition (pedogene- This scenario is completely different from that of the sis). Temperature and moisture affect oxidation or re- Chinese Loess Plateau where there is a positive rela- duction during soil development because of their influ- tionship between magnetic susceptibility and pedogene- ence on pH and Eh. Chinese loess and Siberian/Alaska sis because iron oxides are formed under the dominant loess are the two extreme ends of the relationship be- oxidizing pedogenic environment. tween magnetic χ and paleoclimate. As discussed above, This study demonstrates that although wind velocity there is a critical value of effective moisture distin- may influence the magnetic properties of aeolian dust, guishing two areas of positive and negative correlation. once this material is deposited pedogenesis is the domi- Under this critical value, soil pedogenesis is under con- nant factor affecting variation of magnetic mineralogy ditions of oxidation where extra fine maghemite and and susceptibility. magnetite form and increase χ. Typically, warm moist Magnetic susceptibility has often been measured as a oxidising conditions during interglacials favour the regular parameter in Quaternary paleoenvironmental greatest production of these minerals and therefore a studies; and is regarded as an indicator of summer mon- soon (precipitation). Our research has at least three types positive relationship is found between χ and pedogene- of essential relationship between magnetic susceptibility sis. Such a situation is found in the loess deposits of and pedogenesis: (1) Positive relationship, such as Chi- China and Eastern Europe. If moisture is over the criti- nese loess and East European loess; (2) Negative rela- cal value, the pedogenic environment is toward the re- tion, such as Siberian loess and Alaskan loess; and (3) ducing direction. The iron oxides input by aeolian dust No relation, such as New Zealand loess, Argentina loess are destroyed. Moreover, interglacials favour more satu- and some loess on the edges of the Chinese Loess Pla- rated soils and stronger reducing conditions and lead to teau[23, 24, 37]. This study has also suggested that the sus- stronger destruction of iron oxides (maghemite and ceptibility may be a complicated parameter; its applica- magnetite). At the same time, the more paramagnetic tion as a proxy of summer monsoon strength (precipita- minerals are produced, leading to a negative relationship tion) has its limitations and conditions. Great care between magnetic susceptibility with pedogenesis. This should be take when use it for paleoclimatic study. situation is found in Siberian and Alaskan loess. If moisture swings around the critical value then the loess Xiuming Liu thanks Lanzhou University for a starter grant.

292 LIU XiuMing et al. Sci China Ser D-Earth Sci | Feb. 2008 | vol. 51 | no. 2 | 284-293 中国科技论文在线 http://www.paper.edu.cn

1 Heller F, Liu T S. Magnetostratigraphical dating of loess deposits in 20 Zhu R X, Matasova G, Kazansky A, et al. Rock magnetic record of the China. Nature, 1982, 300: 431―433 last glacial–interglacial cycle from the Kurtak loess section, southern 2 Hovan S A, Rea D K, Pisias NG, et al. A direct link between the loess Siberia. Geophys J Int, 2003, 152: 335―343 and marine 18O records: eolian flux to the north Pacific. Nature, 1989, 21 Begét J, Hawkins D. Influence of orbital parameters on Pleistocene 340: 296―298 loess deposition in central Alaska. Nature, 1989, 337: 151―153 3 Petit J R, Mounier L, Jouzel J, et al. Paleoclimatological and chrono- 22 Begét J, Stone D, Hawkins D. Paleoclimate forcing of magnetic suscep- logical implications of the Vostok core dust record. Nature, 1990, 343: tibility variations in Alaskan loess. Geology, 1990, 18: 40―43 56―58 23 Liu X M, Hesse P, Rolph T, et al. Properties of magnetic mineralogy of 4 Heller F, Chen C D, Beer J, et al. Quantitative estimates of pedogenic Alaskan loess: evidence for pedogenesis. Quat Int, 1999, 62: 92―102 ferromagnetic mineral formation in Chinese loess and palaeoclimatic 24 Liu X M, Hesse P, Begét J, et al. Pedogenic destruction of ferrimag- implications. Earth Planet Sci Lett, 1993, 114: 385―390 netics in Alaskan loess deposits. Australian J Soil Res, 2001, 39: 5 Maher B A, Thompson R, Zhou L P. Spatial and temporal recon- 99―115 structions of changes in the Asian palaeomonsoon―A new mineral 25 Liu X M, Hesse P, Rolph T. Origin of maghaemite in Chinese loess magnetic approach. Earth Planet Sci Lett, 1990, 125: 461―471 deposits: aeolian or pedogenic? Phys Earth Planet Int, 1999, 112: 6 Liu X M, Rolph T, Bloemendal J, et al. Quantitative estimates of pa- 191―201 leoprecipitation at Xifeng in the loess plateau of China. Palaeogeogr 26 Rowan C J, Roberts A P. Magnetite dissolution, diachronous greigite Palaeoclimatol Palaeoecol, 1995, 113: 243―248 formation, and secondary magnetization from pyrite oxidation: Un- 7 Han J M, Lü H Y, Wu N Q, et al. Magnetic susceptibility of modern soils ravelling complex magnetizations in Neogene marine sediments from ― in China and climate conditions. Stud Geophys Geodetica, 1996, 40: New Zealand. Earth Planet Sci Lett, 2006, 241: 119 137 27 Emiroglu S, Rey D, Peterson N. Magnetic properties of sediment in 262―275 the Ría de Arousa (Spain): dissolution of iron oxides and formation of 8 Hao Q Z, Guo Z T. Spatial variations of magnetic susceptibility of Chi- iron sulphides. Physics Chem Earth, 2004, 29: 947―959 nese loess for the last 600 ka: Implications for monsoon evolution. J 28 Hu S, Appel E, Hoffmann V, et al. Gyromagnetic remanence acquired Geophy Res, 2005, 110: B12101 by greigite (Fe S ) during static three-axis alternating field demag- 9 Evans M E, Heller F. Magnetism of loess palaeosol sequences: re- 3 4 netization. Geophys J Int, 1998, 152: 831―842 cent developments. Earth-Science Rev, 2001, 54: 129―144 29 Geiss C E, Banerjee S K. A multi-parameter rock magnetic record of 10 Kukla J G, An Z S. Loess stratigraphy in central China. Palaeogeogr the last glacial-interglacial paleoclimate from south-central Illinois, Palaeoclimatol Palaeoecol, 1989, 72: 203―225 USA. Earth Planet Sci Lett, 1997, 152: 203―216 11 Heller F, Liu T S . Magnetism of Chinese loess deposits. Geophys J R 30 Bina M, Daly L. Mineralogical change and self-reversed magnetiza- astr Soc, 1984, 77: 125―141 tions in pyrrhotite resulting from partial oxidation; geophysical im- 12 Zhou L P, Oldfield F, Wintle A G, et al. Partly pedogenic origin of mag- plications. Phys Earth Planet Int, 1994, 85: 83―99 netic variations in Chinese loess. Nature, 1990, 346: 737―739 31 Zhu R X, Shi C D, Suchy V, et al. The magnetic properties of Czech 13 Liu X M., Liu T S., Heller F. Xu T C, Frequency-dependent suscep- loess and its paleoclimatic significances. Sci China Ser D-Earth Sci, tibility of loess and Quaternary paleoclimate. Quat Sci, 1990, 1: 2001, 31(2): 146―154. 41―50 32 Shi C D, Zhu R X, Suchy V, et al. Identification and origins of sulfides 14 Liu X M, Shaw J, Liu T S, et al. Magnetic mineralogy of Chinese loess in Czech loess. Geophysical Res Lett, 2001, 28(20): 3903―3906 and its significance. Geophys J Inter, 1992, 108: 301―308 33 Lagroix F, Banerjee S K. Paleowind directions from the magnetic 15 Maher B A, Thompson R. Paleoclimatic significance of the mineral fabric of loess pro¢les in central Alaska. Earth Planet Sci Lett, 2002, magnetic record of the Chinese loess and paleosols. Quat Res, 1992, 37: 195: 99―112 155―170 34 Yang S L, Ding Z L. Winter-spring precipitation as the principal con- 16 Chlachula J, Evans M E, Rutter N W. A magnetic investigation of a trol on predominance of C3 plants in Central Asia over the past 1.77 late Quaternary loess/palaeosol record in Siberia. Geophys J Int, 1998, Ma: Evidence from δ13C of loess organic matter in Tajikistan. Pa- 132: 128―132 laeogeogr Palaeoclimatol Palaeoecol, 2003, 235: 330―339 17 Chlachula J. The Siberian loess record and its significance for recon- 35 Vlag P. The magnetic signal in loess-sequences in Alaska. The Insti- struction of Pleistocene climate change in north-central Asia. Quat Sci tute for Quarterly. 1998, 8: 1―7 Rev, 2003, 22: 1879―1906 36 Rosenbaum J G, Reynolds R L, Muhs D R, et al. Geochemical con- 18 Matasova G, Petrovsky E, Jordanova N, et al. Magnetic study of Late straint on the interpretation of magnetic property variations in Pleistocene loess/palaeosol sections from Siberia: palaeoenviron- loess/paleosol sequences from central Alaska. EOS, Transactions, AGU mental implications. Geophys J Int, 2001, 147: 367―380 Fall Meeting, 1997, 78: F170―171 19 Zhu R X, Kazansky A, Matasova G, et al. Rock-magnetic investiga- 37 Sun J M, Liu T S. Multiple origins and interpretations of the magnetic tion of Siberia loess and its implication. Chin Sci Bull, 2000, 45: susceptibility signal in Chinese wind-blown sediments. Earth Planet 2192―2198 Sci Lett, 2000, 180: 287―296

LIU XiuMing et al. Sci China Ser D-Earth Sci | Feb. 2008 | vol. 51 | no. 2 | 284-293 293