A Late Pleistocene palaeoclimate record based on mineral magnetic properties of the entrance facies sediments of Kulna Cave, Czech Republic

Summary

Kulna Cave is located in the Moravian Karst, a well-developed karstic region formed in Devonian limestones in the eastern part of the Czech Republic. The entrance facies sediments in the cave consist of interbedded layers of silts (loess) and clay-rich silts (loam) that were either directly blown into the cave entrance or redeposited in the cave by slope processes during the Last Glacial Stage. The layers of loess and loam overlie fluvial sands and gravels deposited during the Last Interglacial. Previous research at Kulna concentrated on the archaeology, palaeontology and dendrology of these entrance facies sediments. From these data, palaeoenvironmental conditions in the vicinity of the cave were reconstructed. Our results suggest that susceptibility variations and in particular variations in pedogenic susceptibility yield a more detailed record of the palaeoenvironmental conditions at the cave during the Last Glacial Stage.

Magnetic susceptibility (χ) was measured on approximately 700 samples collected throughout three well-studied profiles in the cave entrance. The χ record is well defined and correlates from one profile to another. Mineral magnetic measurements [F_D, ARM/SIRM, S-ratio, χ(T)] suggest that χ variations in the Kulna sediments from the Last Glacial Stage are controlled by the concentration of magnetite and/or maghemite formed during pedogenesis. After the removal of the effects of fine carbonate debris and detrital ferromagnetic minerals on the bulk χ record, we obtained a record of pedogenic susceptibility (χ_p) that serves to quantify the concentration of magnetic minerals formed during pedogenesis. Therefore, χ_p can be thought of as a proxy reflecting the intensity of pedogenesis, which in turn is controlled by climate. Our χ_p record is also in good agreement with the median grain size record of the Kulna sediments (another proxy for climatic change). We suggest that in the case of Kulna, χ_p is more sensitive to climate change than bulk χ.

The Kulna pedogenic susceptibility record shows variations on both long and short timescales. The long-term trends are in good agreement with the deep-sea SPECMAP record, while the short-term oscillations correlate well with rapid changes in the North Atlantic sea surface temperatures. Our results suggest that Central European climate during the Last Glacial Stage was strongly controlled by the sea surface temperatures in the North Atlantic. Short-term warmer events and perhaps higher precipitation over the mid-continent increased the intensity of pedogenesis. Given the number of independent climate proxies determined from the entrance facies of the cave and their high resolution, Kulna is an extremely important site for studying Late Pleistocene climate.

caves, entrance facies, grain size, magnetic susceptibility, palaeoclimate, pedogenesis

Introduction

Magnetic properties of iron oxides found in naturally occurring sediments have been shown to reflect changes in environmental processes operating at the Earth's surface. The most commonly measured magnetic property, mass-specific magnetic susceptibility (χ), is a function of concentration, grain size and mineralogy of magnetic minerals found in the sediments. This dependence has lead researchers to interpret changes in χ ;measured in sedimentary sections in terms of environmental changes occurring at the time of deposition. Attributing environmental significance to χ and other mineral magnetic parameters has spawned a new field of study known as ‘environmental magnetism’. In the last two decades, enormous progress has been made in the successful application of magnetism to palaeoenvironmental and palaeoclimatological problems. Thompson & Oldfield (1986), King & Channell (1991) and Verosub & Roberts (1997), among others, provide excellent reviews of our current state of knowledge. These reviews demonstrate that measurements of χ and other magnetic parameters have a growing role to play in many diverse and important areas of the environmental sciences.

Magnetic minerals are affected by physical and chemical weathering (including pedogenesis), intensity of erosion and the energy of transport of the sediments. Since these processes produce a wide range of sediments, mineral magnetic methods can be utilized in a broad range of environments. In loess–palaeosol sequences, for example, χ proved to be a sensitive detector of the long-term, large-scale terrestrial climatic change (e.g. Kukla et al. 1988; Banerjee 1994; Maher et al. 1994; Heller & Evans 1995). In deep sea sediments, concentration-dependent magnetic parameters have been used to decipher the palaeoclimate record in a very simple, rapid and non-destructive matter (e.g. Robinson 1986; Colin et al. 1998; Arnold et al. 1998). In addition, magnetic properties can be used to discriminate between different fluxes of terrigenous sediments and to detect in situ production (e.g. Tarduno 1994). The realization that χ not only provides a useful parameter for high-resolution core correlation but is also rapidly affected by human impact triggered a wide range of lake-watershed studies (e.g. Oldfield et al. 1979; Banerjee et al. 1981; Peck et al. 1994). Lake sediments became increasingly important for short-term climatic studies as the need to set present-day environmental processes in a longer time perspective became apparent (Thouveny et al. 1994; Snowball 1995).

One sedimentary environment that has not been exploited adequately by the environmental magnetists is the cave environment. The cave environment can be divided from the sedimentological point of view into an entrance facies and an interior facies. The entrance facies includes fine-grained sediments transported from the vicinity of the cave by wind and water and coarser clasts transported into the cave by slope processes. Stratigraphically, the entrance facies is the most valuable section of the cave. The cave entrance contains pollen as well as datable archaeological and palaeontological remains that are protected from surface erosion, weathering and biochemical alteration. The interior facies develops in those parts of the cave that are more remote from the surface. Sedimentary sequences here are often extensive, consisting of coarse gravels and sands of fluvial origin overlain by flood deposits of laminar silts and clays. Due to the dynamic environment of cave interiors, sedimentary sequences often represent a series of depositional and erosional events.

Early magnetic studies conducted in the cave environment were concerned with determining the age of the sediments based on magnetic polarity (Schmidt 1982) or palaeosecular variation (Ellwood 1971; Creer & Kopper 1974, 1976; Papamarinopoulos et al. 1991). These studies also paid considerable attention to characterizing the sedimentary fabric and palaeoflow directions in the interior facies using anisotropy of magnetic susceptibility (Noel & St. Pierre 1984; Noel 1986; Turner & Lyons 1986). However, these studies reported very little about the environmental magnetic properties of cave sediments. To date, only Ellwood et al. (1996) and Sroubek et al. (1996) have attempted environmental magnetic studies in this environment. Ellwood and co-authors produced a palaeoclimate record for the time interval 9000–3500yr BP using magnetic susceptibility measurements on clastic interior facies sediments. Profiles were located in an archaeological excavation pit in Konispol Cave, Albania. The authors correlated χ highs with periods of warm and humid climate as defined from other European regions located nearby. Variations were attributed to the content of pedogenically controlled maghemite that formed outside the cave and was then washed or blown inside. Ellwood (1999) further expanded on his work in Albania and studied several other caves in the Mediterranean region. He correlated the χ record from one cave to another as well as to the climatic record from the Greenland ice core. His new work also stresses the usefulness of the cave environment for studying past climate change. Sroubek et al. (1996) reported preliminary mineral magnetic results from a study of Kulna Cave, Czech Republic. This study yielded a well-defined magnetic susceptibility signal, the peaks in χ being caused by increased concentrations of magnetite and/or maghemite. Climatic change, human occupation, changing source region and depositional processes were suggested as possible candidates controlling the concentration of magnetic minerals.

In this paper we describe a detailed follow-up to our preliminary environmental magnetic study of the clastic sediments in the entrance facies of Kulna Cave (Sroubek et al. 1996). Kulna Cave is located in the Moravian Karst, a well-developed karstic region formed in Devonian limestones in the eastern part of the Czech Republic (Fig.1). The cave has a tunnel-like shape and is 87m long and 8m high with a maximum width of 25m. The long axis of the cave is oriented north–south with large entrances at both ends. Kulna Cave represents a portion of the upper level of a cave system developed in Sloup Valley that has a total length in excess of 6km. The lower cave level lies 65m below the horizontal upper level. Both of these levels are interconnected by several vertical shafts. Some of these shafts are filled with cave deposits while the others are open chasms.

Figure 1

Location of the Moravian Karst in the eastern part of the Czech Republic.

The cave contains approximately 15m of clastic sediments that span nearly the entire Upper Pleistocene. The first excavations in the sedimentary filling of the cave began in 1880 when stone tools and bones of extinct animals were found (Valoch 1988). Systematic research was completed between 1961 and 1976 by scientists from the Moravian Museum in Brno led by Dr Karel Valoch. Valoch (1988) subdivided the entrance facies into 14 recognizable stratigraphic units based on archaeological artefacts, pollen and charcoal analysis, palaeontological findings and several radiocarbon dates. Layers 1–6 were assigned to the Holocene/Late Wurm Glacial Stage, layers 6a–9 to the Early Wurm and layers 10–13 to the Eemian Interglacial Stage. Layer 14, which originated during the Riss Glacial Stage, is presently buried due to partial collapse in the excavation. Layers 1–5 [which were removed due to industrial activity in the cave interior during the Second World War and from the front of the cave during archaeological excavations described by Valoch (1988)] consisted of dark brown clayey silt with varying amounts of limestone clasts. Layer 6 is a yellow brown silt (interpreted to be a loess blown directly into the cave) mostly devoid of clasts, while layer 6a is a yellow–brown clayey silt with a small amount of fine detritus. Layer 7a consists of interbedded layers of yellow–brown and reddish-brown clayey silts, with laminae thicknesses of less than 1cm. Layer 7b is similar to 7a except that it is totally devoid of clasts and laminae thicknesses are much smaller (1–5mm). Pellet sand horizons of colluvial origin (Kukla 1975) are located in the bottom part of this layer. The underlying layers 7c and 7d are both finer grained, with layer 7c being richer in limestone clasts. Layer 8a is a grey–brown, clay-rich silt with variable amounts of clasts. Layer 9 has a similar matrix to 8a but is richer in limestone clasts. Layer 10, a black silt with fine limestone clasts, appeared only in front of the cave and is no longer preserved. Layer 11 is a complex of grey silty clays rich in limestone clasts. Layers 6a–11 are thought to represent deposits of weathered loess (loam) that were transported into the cave by meteoric waters during precipitation-rich periods from slopes surrounding the cave entrances. A retreating entrance also contributed limestone blocks to these sediment layers. Layer 12 is a dark brown fluvial sediment consisting of sandy gravels. Layer 13 forms a lens of sands and clayey silts within layer 12. Layers 12 and 13 were deposited during the Last Interglacial by streams flowing through Sloup Valley that sank into vertical cavities connecting Kulna Cave with the lower-cave level.

Valoch (1992) has correlated the Kulna stratigraphy with the oxygen isotope stages (OIS) of Labeyrie (1984), thereby tying the Central European climatic record to variations in the global ice volume. Table1 and Fig.2 summarize the findings of Valoch (1988, 1992) in the layers from the Last Glacial and Last Interglacial.

Table 1

Summary of the results from the previous study of the Kulna Cave (modified after Valoch 1988, 1992). The unmarked ages are uncalibrated conventional radiocarbon dates.

Figure 2

Simplified stratigraphic profile of the Late Pleistocene sequence in the entrance of Kulna Cave. Ages are radiocarbon dates from Valoch (1988), except for the 50ka age, which is an ESR date from Rink et al. (996). The correlation with the deep-sea stable oxygen isotope record is from Valoch (1992). Oxygen isotope age model is from Martinson et al. (1987).

In our preliminary study (Sroubek et al. 1996), we sampled three profiles (Fig.3) described by Valoch (1988) located in different parts of the cave entrance (Fig.4) to determine whether the magnetic susceptibility signal showed similar patterns. From each profile, we collected 8cm³ samples of the fine-grained matrix at 10cm vertical spacing. Measurements of mass-specific magnetic susceptibility (χ) indicated that sediments in the cave yielded a well-defined signal that correlated from one profile to another (see Fig.9 of Sroubek et al. 1996). Additional magnetic parameters suggested that the variation in χ is controlled mostly by the concentration of magnetite and/or maghemite rather than by mineralogical or grain size changes.

Figure 3

Sketch of (a) profile 4′, (b) profile 6, and (c) profile 7 in the entrance of Kulna Cave. Squares indicate locations of preliminary samples; bars indicate locations of detailed sampling (sample spacing 2.7cm) as discussed in the text. The modern cultural layer shown in profile 7 (see legend) includes a support and a metal ramp constructed for easier access to the profile.

Figure 4

Sketch of the southern entrance and central part of Kulna Cave. Pits excavated by Valoch and co-workers are represented by hatched polygons. Thicker lines (labelled 4′, 6 and 7) represent locations of sample profiles.

Figure 9

Directions of the three principal axes of magnetic susceptibility for sediments from layers 7b−11 of Kulna Cave. Squares represent maximum axes, triangles intermediate axes and circles minimum axes. The larger symbols represent the means and the ellipses indicate the 95 per cent level of confidence of the mean. The orientations of the mean directions are as follows. Maximum: D=143°, I=0°; intermediate: D=233°, I=4°; minimum: D=52°, I=86°.

Sampling and laboratory methods

Our present, more detailed work in Kulna Cave focuses on obtaining a high-resolution mineral magnetic record and on interpreting the palaeoclimatic significance of this record. To this end, we collected close to 700 unoriented 8cm³ (2cm cube) sample from the same profiles sampled for our preliminary work (Fig.3). All the layers were sampled continuously (distance between centres of samples on average 2.5cm) except for layers 9 and 11, which were sampled at a 10cm interval in order to preserve the integrity of the profile and to avoid common limestone clasts. Note that in the sampled profile 7, layer 13 overlies layer 12.

χ of all samples from each of the profiles was measured on a Kappabridge KLY-2 instrument. The magnetic susceptibility measurements were complemented by the following rock magnetic parameters. Paramagnetic susceptibility (χ_para) was determined on three–four samples from each layer from the high-field segments (above 300mT) of the hysteresis loops as measured on a variable-frequency translation balance (VFTB). Samples from certain layers contained abundant carbonate clasts. To remove the effect of carbonates on magnetic susceptibility, all samples were treated with acetic acid to dissolve the carbonate as the final procedure in our laboratory analysis. Frequency dependence of magnetic susceptibility (F_D) was calculated from the average of three repeated measurements obtained on a Bartington Instruments MS-2 susceptibility meter at 470 and 4700Hz. Saturation isothermal remanent magnetization was determined for every fourth sample at 1T (SIRM, IRM_1000mT) and at −300mT (IRM_−300mT) on a Sapphire Instruments SI-6 pulse magnetizer. The ratio IRM_−300mT/IRM_1000mT was used to obtain the S-parameter. The anhysteretic remanent magnetization (ARM), determined again on every fourth sample using the Sapphire Instruments SI-4 AF demagnetizer (biasing field 0.05mT in an alternating field of 1T), was used to calculate the magnetic grain size-dependent ratio ARM/SIRM. All remanence parameters (IRMs and ARMs) were measured on a Schonstedt SSM-2 spinner magnetometer. The behaviour of magnetic susceptibility was measured against increasing temperature on the Kappabridge KLY–2 equipped with a furnace CS-2 on two samples from every layer. The heating (from 20 to 700°C with a step of 3°C) was carried out in an argon atmosphere in order to prevent oxidation of the sample. Anisotropy of magnetic susceptibility (AMS) was measured on 15 oriented cylindrical samples (6cm³) from each layer except for layer 6a, for which three samples were measured. This layer was no longer accessible after the manager of Kulna (Moravsky Kras Co.) opened the cave to the public. The AMS tensor for each sample was calculated from 15 individual measurements made on the Kappabridge KLY-2 as prescribed by methods developed by Jelinek (1978).

To complement our magnetic measurements we measured the grain size of the fine-grained matrix of our samples. Grain size distributions (0.7–700µm) were measured on 150mg subsamples using a 20 channel Leeds and Northrup MICROTRAC II laser particle size analyser. The morphology of quartz grains (quartz exoscopy) was studied in order to determine the mode of transport and post-depositional changes. Five grains from each layer were analysed under high magnification using the JOEL jsm-35 C electron microscope.

To compare our work in Kulna with other similar surficial deposits we collected 10 samples from loess and 10 samples from palaeosols from the Cerveny Kopec (Red Hill) outcrop in Brno and 10 samples from the fluvially redeposited loess-like sediments in a river terrace of Bila Voda Stream near Holstejn village in the Moravian Karst. Magnetic susceptibility, frequency dependence of magnetic susceptibility and grain size distribution were measured on these samples.

Laboratory results

The detailed susceptibility records from profiles in the entrance of the Kulna Cave show, as expected, a good correlation (Fig.5), as was also the case with our preliminary work. The similarity of our susceptibility records from profile to profile also argues against any magnetic contamination by industrial activity in the cave during the Second World War, which occurred before Valoch's (1988) archaeological excavations exposed the profiles we sampled. In the two exposures of layer 6a, χ shows low values between 2 and 3×10⁻⁷m³kg⁻¹. A rise in susceptibility up to 4×10⁻⁷m³kg⁻¹ can be observed in all three exposures of layer 7a. Layer 7b contains two distinct peaks in the lower half of the layer reaching 6×10⁻⁷m³kg⁻¹. Except for profile 4′, χ drops in all profiles across layer 7c and the top of layer 7d to 3×10⁻⁷m³kg⁻¹. In the lower part of layer 7d χ again increases to 5×10⁻⁷m³kg⁻¹. Layers 8a, 9 and 11 show fairly rapid changes in χ-values. Finally, the sediments of layers 12 and 13 at the base of the profile have the lowest χ-values, around 1×10⁻⁷m³kg⁻¹.

Figure 5

Detailed record of the magnetic susceptibility measured throughout the main entrance of Kulna Cave. Profile and sampling locations are shown in Fig.4.

Since χ trends in the same layer sampled from different profiles are similar, data from these profiles were spliced together to form a composite record (Fig.6a) summarizing χ variations in layers 6a–13 of the entrance facies in Kulna Cave. The χ data shown in Fig.6(a) come from layer 6a in profile 4, layers 7a, 7b, 7c, 7d and 8a in profile 6, and layers 9, 11, 12 and 13 in profile 7. The selected layers represent the maximum thickness of that unit found in the cave profiles.

Figure 6

Magnetic parameter data from the composite profile constructed from the individual profiles sampled in the entrance facies of Kulna Cave. (a) χ, (b) F_D, (c) ARM/SIRM, (d) S-ratio, (e) carbonate content in weight per cent, and (f) χ_cc.

The composite record of χ shows in general a good agreement with the climatic interpretation of Valoch (1988) and the conclusions of Ellwood et al. (1996; Fig.7), that is, the highest χ-values are associated with layers interpreted to have been deposited during a warmer climate. However, several discrepancies are apparent. For example, layers 7c and 7d, which according to Valoch (1992) originated during relatively warmer OIS 5a and OIS 5b, have lower χ-values than the overlying layer 7b deposited during stadial conditions (OIS 4). Similarly, layers 9, 11, 12 and 13, deposited during the near-optimal conditions of OIS 5d and 5e as suggested by Valoch (1992), show generally very low χ-values.

Figure 7

High-temperature susceptibility measurements of samples from the Kulna Cave sediments. Samples were heated in argon atmosphere; only the heating curves are shown.

To explain the discrepancies between the χ record and the general climate reconstruction data discussed above we have undertaken the following experiments to determine whether χ variations are caused by changes in concentration, grain size or magnetic mineralogy: (i) measurements of frequency dependence of χ (F_D, Fig.6b) in order to estimate the concentration of ferromagnetic grains near the SD/SP boundary, (ii) measurements of the ARM/SIRM ratio (Fig.6c) in order to characterize variations in grain size of the remanence-carrying grains (SD and larger; larger values of ARM/SIRM indicate smaller magnetic grain size), (iii) measurements of the S-ratio (Fig.6d) in order to quantify the ratio between high- and low-coercivity minerals, (iv) measurements of paramagnetic susceptibility quantifying the contribution of paramagnetic minerals to the bulk χ, and (v) measurements of χ with increasing temperature (Fig.7) in order to determine the magnetic minerals present.

The record of F_D shows a gradual increase down-section in the upper part of the profile (layers 6a, 7a) from values as low as 2 per cent up to approximately 9 per cent. Throughout most of the profile (layers 8a–11) the F_D values are between 9 and 11 per cent. In the two lowermost layers (13 and 12) F_D values are once again quite low, typically below 5 per cent. The ARM/SIRM record has a similar trend to the F_D data. The ARM/SIRM record shows a rather scattered increase down-section throughout layer 6a, indicating a decrease in magnetic grain size. Throughout layers 8a–11 the ARM/SIRM values shows minimum variability, but decrease towards the base of the profile in layers 13 and 12. These layers thus contain larger magnetic grains. The S-ratio is nearly constant (between 0.85 and 0.95) in layers 6a–11, then drops rapidly at the boundary between layers 11 and 13 and is again nearly constant throughout layers 13 and 12 (between 0.75 and 0.8). The S-ratio data indicate that the upper layers contain an increased concentration of low-coercivity magnetic minerals compared to the bottom two layers. Paramagnetic susceptibility (data not presented) contributes approximately 20 per cent to the bulk χ signal in layers 6a–11 and nearly 40 per cent in layers 13 and 12. High-temperature susceptibility curves for layers 7b and 7d show a very similar trend (Fig.7). Both curves show nearly constant values of χ up to about 200°C, after which they increase to a maximum slightly below 300°C and then decrease until about 400°C. Above 400°C χ rises rapidly until 450°C and then drops to a constant background value. The first rise in the record probably indicates the transition of Fe-hydroxides to maghemite, the χ loss between 300 and 400°C is attributed to maghemite converting to haematite, the second χ rise suggests reduction of haematite to magnetite and the χ loss around 580°C confirms the presence of magnetite and/or a phase converting to magnetite. The sample from layer 6a shows a similar trend to the two previously discussed samples except for the absence of the first peak at approximately 300°C (Fig.7) The second peak is present in this sample, but is less pronounced. Layer 6a thus probably contains no (or a minimum amount of) Fe-hydroxide and less haematite. Our preliminary results on IRM acquisition and thermal demagnetization of SIRM presented in Figs5 and 6 of Sroubek et al. (1996) also suggest the presence of maghemite or low-coercivity magnetite in all the layers of the entrance facies of Kulna.

The mineral magnetic properties of layer 6a suggest that the low χ-values are caused by a lack of grains near the SP/SD boundary. In layers 7b, 7c, 7d, 8a, 9 and 11 the content of the grains near the SP/SD boundary is higher and χ variations are probably controlled by changes in concentration of these fine grains. Layer 7a appears to be a transitional layer. In its upper part the magnetic grain size and concentration of grains near the SP/SD boundary is low and similar to layer 6a. Throughout layer 7a the magnetic grain size decreases and at its base is similar to the underlying layer 7b. Magnetic mineralogy throughout layers 6a–11 does not show significant variations except for the lack of Fe hydroxides in layer 6a. The sandy gravels of layers 13 and 12 show a distinctly different magnetic signature as indicated by the increased presence of paramagnetic and high-coercivity ferromagnetic minerals as well as by the increase in magnetic grain size.

Mineral magnetic properties of layers 6a–11 are very similar to those of loess/palaeosol sequences exposed at Cerveny Kopec. Low values of F_D in layers 6a and 7a are similar to those we measured on unweathered loess at Cerveny Kopec (F_D=3–5 per cent), whereas the high values in layers 7b–11 correspond well with palaeosol samples from Cerveny Kopec (F_D=8–11 per cent). Similar elevated values of F_D are common in palaeosols interbedded in loess sequences from around the world (e.g. Evans & Heller 1994; Eyre & Shaw 1994). In addition, the high-temperature susceptibility curve for layer 6a is similar to the results of Oches & Banerjee (1996) measured on loess samples from the classic profile at Dolni Vestonice, whereas the high-temperature susceptibility curves for layers 7b and 7d are similar to results on the parabraunerde soils from Dolni Vestonice.

The above discussion suggests that sediments in layers 7a–11 underwent pedogenic weathering. This is also supported by the analysis of quartz grain surfaces from these layers. Pedogenic dissolution as well as in situ production of new phases is apparent from the morphology of quartz grain surfaces (Fig.8). As seen in Fig.8, both dissolution features and newly precipitated silica on the surface of a wind-blown silt-sized quartz grain are apparent.

Figure 8

Electron microscope images of quartz grains from layer 7b of the entrance facies sediments of Kulna Cave. The surface features yield evidence of the mode of transport and of post-sedimentary processes.

The pedogenic weathering of the sediments probably occurred either in the source area or more probably in the vicinity of the cave immediately prior to transport into the cave entrance. In order to obtain information on the mode of deposition we conducted AMS measurements. AMS data presented in Fig.9 suggest that settling velocities were low as the foliation plane for these sediments is near-horizontal and the degree of anisotropy is minimal (below 3 per cent). These data are typical for an environment in which the anisotropy is controlled by the compaction process rather than by settling energy or surface topography (Tarling & Hrouda 1993). The direction of the mean principal anisotropy axis, pointing NW–SE, suggests that the sediments were transported into the cave through the southern cave entrance. Directional AMS data for each layer indicate that at the 95 per cent confidence level there is no difference in the direction of transport or sedimentary fabric between the layers.

The similarity between the magnetic properties of the sediments deposited in layers 6a–11 and those of the loess/palaeosol sequences suggests that the χ variations throughout layers 6a–11 should be a good proxy for climate during the Last Glacial. The reason why we observe discrepancies between the climatic record of Valoch (1988, 1992) and our χ data is probably due to the fact that magnetic susceptibility measured in Kulna is not only a measure of the concentration of pedogenic ferromagnetic minerals but is also affected by both paramagnetic (mostly clays) and diamagnetic (quartz and carbonates) minerals. These contributions, if significant, must be subtracted from the bulk χ to obtain a true pedogenic signal. The concentration of paramagnetic minerals in Kulna Cave is fairly constant (∼ 20 per cent) throughout the layers and thus does not contribute significantly to the susceptibility variations shown in Fig.8(a). On the other hand, the variable amount of carbonate clasts (Fig.6e) affects the χ signal. In layer 6a the carbonate content is approximately 10 per cent and in layers 7d, 9 and 11 it varies between 10 and 50 per cent. In all the other layers the carbonate clast content is mostly below 5 per cent. The magnetic susceptibility signal after the subtraction of the diamagnetic carbonate contribution (carbonate-corrected susceptibility, χ_cc) as shown in Fig.6(f) has a similar character to the bulk χ record in Fig.6(a) but shows significantly higher values than the χ record in the carbonate-rich layers.

The χ_cc record is composed of a pedogenic component created by authigenic grains near the SP/SD boundary and a background component originating from larger detrital ferromagnetic grains. Forster et al. (1994) suggest that pedogenic and background susceptibilities can be separated utilizing the delχ versus χ_cc plot, where delχ is the difference in magnetic susceptibility measured at two frequencies. The assumption behind this test is that pedogenic susceptibility (χ_p) is carried by grains near the SD/SP boundary and the larger grains, which show no frequency dependence of χ, are responsible for the background susceptibility (χ_b). χ_b is defined as the intercept between the best-fit line to the data points on the delχ versus χ_cc plot and the χ_cc axis. The pedogenic component (χ_p) is the difference χ_cc−χ_b, and F_c (per cent), the ‘collective true frequency dependence’ characterizing the grain size distribution of the SP fraction, is the slope of the best-fit line to the delχ versus χ_cc plot. In determining χ_b and χ_p in the Kulna sediments we followed the approach of Forster et al. (1994) and constructed delχ versus χ_cc plots for each layer in Kulna (Fig.10). The best-fit lines suggest that layers 7b, 7c, 7d and 8a have both very similar intercepts or χ_b (between 1 and 1.4×10⁻⁷m³kg⁻¹) and slopes or collective F_c (around 12 per cent). Layers 9 and 11 have a somewhat higher χ_b=2.41× 10⁻⁷m³kg⁻¹ but exhibit the same F_c. Layer 6a is distinctly different, with an intermediate χ_b=2.04×10⁻⁷m³kg⁻¹) and a very high F_c (21 per cent). Layer 7a appears to be a transition between 6a and the underlying layers with χ_b=1.63× 10⁻⁷m³kg⁻¹ and F_c=16 per cent. The results from Kulna's layer 6a are similar to unweathered loess beds around Brno, which yield χ_b values between 1.5 and 2.2×10⁻⁷m³kg⁻¹ (Forster et al. 1996). The rest of the layers in Kulna have characteristics that correspond well to the Late Pleistocene loess/palaeosol sequences of Southern Moravia (χ_b=0.8–1.1× 10⁻⁷m³kg⁻¹ and F_c around 13 per cent, Forster et al. 1996).

Figure 10

Plot of magnetic susceptibility after removal of carbonate debris (χ_cc) versus the difference between low- and high-frequency magnetic susceptibility (χ_lf−χ_hf) for individual sedimentary layers deposited in the entrance of Kulna Cave. The best-fit line through the data points yields the background susceptibility (χ_b), equal to the intercept χ_cc axis and the collective true frequency dependence (F_C) given by the slope of the best-fit line. R² is the coefficient of determination measuring the quality of the fit of the line to the data points. For further explanation see text.

Utilizing the χ_b values for each layer, we calculated χ_p for each sample in our composite profile (Fig.11). The record of χ_p shows lower values than the χ_cc record (Fig.6f). The difference between χ_p and χ_cc is most significant for layers 6a and 7a, which consist mostly of unweathered loess where the pedogenic component is minimal. Also in layers 9 and 11 the χ_p values are significantly lower than in the χ_cc record. One possible explanation is that during intense pedogenesis the newly formed magnetic grains can grow to larger sizes than the SP/SD boundary and then behave as if they were part of the background signal. If this explanation is correct the method of Forster et al. (1996) must be applied and interpreted with greater care.

Figure 11

Comparison of pedogenic susceptibility (χ_p) and median grain size measured through our composite profile in the entrance of Kulna Cave. For details on obtaining χ_p see text and Fig.12.

Grain size measurements on the sedimentary matrix in Kulna were conducted in order to provide independent climatic proxy data that could be compared with the χ_p record. Pure loess is typically well sorted, with up to 70 per cent of the grains in the range between 10 and 50µm and has a positively skewed size distribution (Kukal 1971; Pecsi 1990). During pedogenesis, however, grain size distribution of weathered loess becomes bimodal and the mean grain size decreases with the formation of clay minerals (Martini & Chesworth 1992). Variations in mean grain size and other parameters characterizing grain size distribution thus can be used as climate proxies. For example, Ding et al. (1998) correlated grain size variations in Chinese loess with the Dansgaard–Oeschger events and Chen et al. (1998) used the silt/clay ratio as a measure of the intensity of the monsoons. On a more local scale, our grain size measurements from the loess/palaeosol complex at Cerveny Kopec (Figs12a and b), 30km SW of Kulna, show changes in the character of the grain size distribution that can also be attributed to climatic change. Pristine loess from stadials (Fig.12a) is positively skewed with median grain size varying between 20 and 35µm, and the palaeosols formed during interstadials (Fig.12b) are bimodal with median grain sizes as low as 9–14µm.

Figure 12

Size distribution of the fine-grained matrix in the entrance facies of Kulna Cave compared to sediments of similar origin. (a) Loess samples from Red Hill (Cerveny Kopec); (b) palaeosol sediments from Red Hill (Cerveny Kopec); (c) samples of fluvial sediments deposited by Bila Voda River; (d–j) samples from layers in the entrance of Kulna Cave.

The grain size distribution of the matrix of most Kulna sediments (Figs12d–j) is in all cases bimodal, with the silt peak between 30 and 60µm and the clay peak between 3 and 5µm. We attribute the silt peak to the presence of windblown particles and the clay peak to grains formed during the pedogenic weathering. The ratio of the clay peak to the silt peak therefore reflects the intensity of the pedogenic process prior to deposition in the cave. In layer 6a (Fig.12d), the grain size distribution is very similar to the loess (Fig.12a); the silt peak strongly dominates suggesting the presence of mostly aeolian particles. In layer 7a the clay peak becomes more prominent as silty clay (loam) laminae are interbedded with silty (loess) layers. In layer 7b, samples with both the clay peak dominating and the silt peak dominating can be observed. Layer 7b consists of thin laminae of strongly weathered loess (loam), responsible for the high relative concentration of clay particles, and loess laminae that are composed mostly of silt-sized grains. In layers 7d, 8a, 9 and 11 the clay peak dominates over the silt peak, suggesting intensive weathering. In some cases the silt peak flattens into a shoulder, probably due to the presence of fine carbonate detritus. The size distribution of layers 12 and 13 is very similar to surficial fluvial deposits (Fig.12c) from the nearby Bila Voda River.

The ratio between the silt peak (loess) and the clay peak (loam) is well reflected in the median grain size record (Md, Fig.11). The nearly pristine loess in layer 6a thus has maximum grain size, with the rapid variations in the Md probably recording wind intensities. In layer 7a the Md shows an overall decrease, representing a more intense weathering process and the increased presence of clay-rich materials (loam). The rapid variation in Md in layer 7b reflects the presence of alternating clay-rich loam (low-Md) and loess (high-Md) laminae. In layers 7d, 8a, 9 and 11 Md stays nearly constant at its minimum values, suggesting that only clay-rich materials (loam) are present. Finally, the fluvial sediments in layers 13 and 12 have overall slightly higher Md. The rapid rise and drop in layer 13 reflects the presence of a sandy lens, reflecting a local increase in the energy of the depositional environment. A general inverse correlation between the two palaeoclimate proxies Md and χ_p is apparent in layers 6a–7c. The variations in Md and χ_p in the massive layer 6a and thickly laminated layer 7a show a high degree of inverse correlation, which can also be seen in the fine structure of both records. In layer 7b, this detailed correlation is not apparent, mostly because of the different sampling methods used for magnetic and grain size measurements. In layer 7a the thickness of laminae is roughly equal to the size of the sampling box, but in the finely laminated layer 7b, the magnetic susceptibility sample reflects an averaged signal of several laminae, whereas the much smaller grain size sample is lithologically homogeneous. However, when samples from individual laminae are compared, the correlation between magnetic susceptibility and median grain size becomes apparent. For example, in the laminae of layer 7b there is a direct correlation between the median grain size and χ. The yellow–brown laminae of aeolian origin have very low χ (2–3×10⁻⁷m³kg⁻¹) and very fine median grain sizes (∼7µm), whereas the adjacent reddish laminae show significantly higher χ (∼6×10⁻⁷m³kg⁻¹), coarser median grain sizes (∼14µm) and a bimodal grain size distribution.

In layers 7d, 9 and 11 the correlation between Md and χ_p is not as pronounced as in the upper part of the profile. The Md values are nearly constant, while χ_p shows distinct variations. The likely explanation is that the measured Md is not as sensitive to the intensity of the pedogenic process as χ_p is. The grain size decrease probably reaches a certain minimum threshold as the lower limit of the grain size measuring instrument is 0.7µm. χ_p, on the other hand, is extremely sensitive to changes in concentration of particles around 0.03µm that cannot be detected by the grain size measurements. This conclusion is supported by the good inverse correlation between Md and the magnetic grain size of remanence-bearing magnetic minerals (ARM/SIRM). The ARM/SIRM, which is sensitive to grain size changes in a comparable range to Md, also shows a nearly constant progression throughout layers 7d–11 and increases significantly only in layers 6a, 7a, 13 and 12.

Discussion

Results from the sediments in Kulna Cave suggest that pedogenic magnetic susceptibility (χ_p) and grain size variation are sensitive indicators of the intensity of the pedogenic processes responsible for producing these variations. Since pedogenesis is in part controlled by climate, our susceptibility and grain size variations may be a proxy for climate change. Fig.13 compares the χ_p record with the SPECMAP oxygen isotope record (Imbrie et al. 1984) and with the North Atlantic sea surface temperature data derived from core K708-1 (Ruddiman 1987). The SPECMAP oxygen isotope record represents general climatic trends as the record was created by averaging worldwide data. Temperature data from core K708-1, located at 50 °N, are, however, probably a better proxy of climatic variations at Kulna Cave during the Late Pleistocene as North Atlantic ocean atmosphere interactions have a dominant impact on the climate of Central Europe (Bradley 1999).

Figure 13

Comparison of the North Atlantic sea surface temperature record with the pedogenic susceptibility measured in the sediments of Kulna Cave and the average deep-sea oxygen isotope record. Light-shaded bands indicate correlation of cold episodes; dark-shaded bands indicate correlation of warm episodes. (a) SST_w: winter sea surface temperature recorded from core K708-1 located approximately 50°N, 24°W (after Ruddiman 1987); (b) χ_p: pedogenic magnetic susceptibility measured through a composite profile in the entrance of the cave; (c) smoothed and stacked record of δ¹⁸O variations obtained from five cores and measured on shallow-dwelling planktonic foraminifera (Imbrie et al. 1984).

The correlation between our χ_p record and the established Late Pleistocene climatic records is based on rather limited age control for the Kulna sediments. Radiocarbon ages are available for layers 6a and 7a, but two of these are uncalibrated conventional dates (layer 6a: 21ka; layer 7a: 39ka) determined in the late 1960s and early 1970s, while the third is a more recent date (layer 7a: 46ka) determined on a sample collected during the excavation of the cave (c. 1961–1976). The climatic optimum (5e) for layers 13 and 12 is assumed based on the change in depositional conditions to a fluvial environment. ESR dates for layer 7a (50ka) are in good agreement with the radiocarbon dates from this layer. However, the 69ka ESR date for layer 9 appears to be too young. A total of six layers from the cave were dated by the ESR technique, out of which only two did not violate the principal of superposition. The possibility of a systematic error is also probable for the ages reported for layers 7a and 9. Therefore, we do not put any weight on the ESR dates. Most of our age control is thus based on matching the χ_p record over the central part of the profile (layers 7b to 11) with the K708-1 sea surface temperature data. Our χ_p record best correlates with the K708-1 sea surface data if layer 7b was deposited between 30 and 50ka. This correlation indicates that the radiocarbon ages from layer 7a should be younger by ∼10ka. Better radiocarbon dates are clearly needed, but given the remarkable similarity between our χ_p and the sea surface temperature data from core K708-1 better radiocarbon dates would only help to fine-tune this correlation

Beginning with the onset of the Last Glacial, the χ_p determined from layers 6–11 shows an overall good agreement with the SPECMAP record and a very high degree of correlation with the North Atlantic sea surface temperatures. This correlation does not exist at the base of the section (layers 13 and 12) because of the different lithological character of these sediments. Fluvial sediments deposited in layers 13 and 12 show a totally different mineral magnetic signature, yielding low χ_p values related to the change in source area of these deposits and their depositional environment.

Based on our results and on the correlation of the χ_p record with the SPECMAP and North Atlantic sea surface temperature records we propose the following reconstruction of the climatic conditions in the vicinity of Kulna Cave. During the climatic optimum (OIS 5e) a stream flowing through the cave was depositing sands and gravels (layers 12 and 13). The source of these sediments was Lower Carboniferous Greywackes, which crop out only 2km north and upstream of the cave. After the stream ceased flowing into the cave, deposition consisted of weathered loess (loam) together with limestone clasts being transported into the cave from the southern entrance. The heavy mineral assemblage of these materials points to their origin as being from granitoids of the Brno Massif (Kvitkova 1999) outcropping only 5km to the west of the cave. The sand- and silt-sized particles were transported mostly by the eastward-blowing winds (Kukla 1975) and mantled the slopes of the Sloup Valley. Conditions at the beginning of the Last Glacial were mild, leading to intensive pedogenesis of the sediments in the vicinity of the cave, prior to their transport into the cave entrance. The varying χ_p throughout layers 7d, 8a, 9 and 11 reflects the slight climatic oscillations during the OIS substages 5a, 5b, 5c and 5d (see Fig.13). Climatic deterioration and lower productivity of pedogenic ferromagnetic minerals during OIS 4 led to lower values of χ_p near the top of layer 7d and in layer 7c. Slight climatic improvement during the Pleniglacial (OIS 3) is reflected in more intensive pedogenesis during deposition of layer 7b. In fact, several short-lived episodes of significant climatic amelioration are represented by the three χ_p peaks in this layer. As the climatic conditions during the Late Glacial (OIS 2) were deteriorating, the pedogenic activity was decreasing and windblown sediments were deposited directly into the cave entrance. At first the deposition of loam and loess alternated, as reflected by the rapidly changing χ_p in layer 7a, but eventually (in layer 6a) loam deposition ceased altogether. What limited pedogenesis occurred resulted in a narrow size distribution of grains around the SD/SP boundary (Eyre 1997), yielding extreme values of F_c (Fig.10).

Our Kulna record as well as the single-core North Atlantic sea surface temperature data show significantly more variability than the SPECMAP record. Also, many other marine oxygen isotopic records as well as other marine proxies (e.g. foraminifera, debris content) point to a significantly higher degree of environmental change (Ruddiman 1987; Heinrich 1988) than the three-fold sequence (OIS 2,3,4) of the SPECMAP. Periods of relatively short-termed warm episodes during OIS 4 are not restricted only to the deep-sea record. Well-defined intervals of relatively mild climatic conditions during the middle and later parts of the Last Glacial have been identified in the oxygen isotope profiles from the Greenland ice sheet (Dansgaard et al. 1982). The millennia-scale ‘Dansgaard–Oeschger’ interstadial events appear to have involved a shift in climate between warm and cold stages of around 7°C. The close correspondence between the ocean and ice-core records suggests that the atmosphere and ocean surface were a coupled system repeatedly undergoing a massive reorganization. However, relating these major climatic changes to the terrestrial record has proven to be more difficult because (i) the atmospheric circulation providing the link between oceans and continents is not yet fully understood, and (ii) many terrestrial proxies are not sufficiently sensitive to record such low-amplitude and short-lived climatic events (Lowe & Walker 1997). Global circulation models (GCM) (e.g. Wright 1993; Broecker et al. 1990) yield more insight into the first point, attempting to establish dynamic links between atmospheric circulation and forcing mechanisms such as the melting of ice sheets, oceanic circulation, vegetation cover or the presence of dust in the atmosphere. Recently, Ruddiman et al. (1989) and Raymo et al. (1990) discovered major feedback mechanisms between melting of ice sheets, fluxes of freshwater into the North Atlantic and the North Atlantic Deep Water (NADW) circulation. According to Huntley & Prentice (1993) these processes are tied to the continental climate in Europe and directly affect the biomass both in the ocean and on the land.

Banerjee (1997) approaches the question of recording short-term changes in the terrestrial record and suggests that ferromagnetic minerals forming in soils typically respond faster to climatic change than most of the other proxies. However, the palaeoclimate record from the loess/palaeosol sequences is often lost due to later processes, in particular due to diagenetic changes occurring in polygenetic soils (Oches & Banerjee 1996). One has to bear in mind that mineral magnetic properties reflect not only processes creating ferromagnetic minerals, but also those destroying them. We suggest that the χ_p measured in Kulna Cave records rapid climatic variations in OIS 3 because of the protective character of the cave. The soil weathered on the surface was transported into the cave where it was not exposed to weathering agents. Thus the cave preserves or fossilizes any environmental impact that modifies the sediment outside the cave.

Our interpretation of the climatic conditions near Kulna show an overall good agreement with that of Valoch (1988, 1992). The only major discrepancy exists in the interpretation of layer 7c. According to Valoch, the thin layer 7c originated during a milder interstadial climate and is correlated with OIS 5a (see Table1). Our results, on the other hand, suggest that layer 7c and the top of layer 7d witnessed a cooling episode that correlates with OIS 4. Valoch's interpretation is based solely on large and small fauna. However, since gastropods and plant remains used in his climate reconstruction are found in limited quantities, a definitive correlation to the OIS record is not possible. This interpretation is also complicated by the possible multiple origin of the fossil material, yielding biased information about the large fauna in the vicinity of the cave. In addition, newer research indicates that some species found in layer 7c (e.g. Cervus elaphus, red deer) are associated with a variety of environmental conditions ranging from oak woods to full glacial tundra (Stuart 1982).

Conclusions

• 1

  The results from Kulna suggest that cave sediments can yield a detailed climatic record of changing conditions during the Last Glacial/Interglacial. However, the transport, sedimentary and post-depositional processes in the cave environment are complex and lead to lithological and textural variations in the strata. Prior to attempting a climatic reconstruction, a high degree of understanding of the depositional history must be obtained.

• 2

  We suggest that the pedogenic susceptibility χ_p measured in the cave sediments is a better climate proxy than χ. χ_p is a sensitive indicator of the intensity of pedogenic weathering and thus records changing climatic conditions in the vicinity of the cave. χ_p correlates well with another proxy for pedogenic weathering—the median grain size of the matrix of Kulna sediments.

• 3

  The χ_p record from Kulna shows a general agreement with the SPECMAP record over the time period 110–15ka. The detailed structure of the χ_p record for this time interval is nearly identical to the sea surface temperature data from the North Atlantic. The high degree of correlation suggests that during the Last Glacial Stage, Central European climate was strongly controlled by the sea surface temperatures in the North Atlantic. Short-term warmer events and perhaps higher precipitation over the mid-continent increased the intensity of pedogenic weathering.

• 4

  The proposed climatic interpretation is in broad agreement with the interpretation of Valoch (1988, 1992), which was based on archaeological, zoo- and phytopalaeontological observations. The large amount of data collected from Kulna allows a multifaceted view of the conditions around the cave as well as of the processes that lead to sediment deposition. Kulna is the first site in the Czech Republic where one single profile has yielded such a profusion of data. The amount of independent climatic information and the high resolution of the record make Kulna an extremely important site for studying the Late Pleistocene climate.

Acknowledgments

M. Renn assisted in the laboratory. We wish to thank W. Lowrie and F. Heller for making the palaeomagnetic laboratory at ETH available to us, M. Krs and P. Pruner at the Geological Institute of the Czech Academy of Sciences for the use of their Kappabridge, and the personnel of the IRM in Minneapolis. Two anonymous reviews helped make this a better manuscript. The reviewers are sincerely thanked. This work was support by National Science Foundation grants INT-9507137 and EAR-9705718, and a grant from the Czech/US Joint Board for Science and Technology (Project no. 95051).