Environmental Reconstruction from the Identification of Magnetic Minerals in the Upper Sedimentary Infill of the Gran Dolina Cave (Burgos, Spain)

Abstract

The cave system in the Sierra de Atapuerca holds one of the most important archaeological sites for the understanding of early human occupation in Europe. Among the different cavities and galleries, the Gran Dolina cave yielded a new hominin species coined as Homo antecessor of an Early Pleistocene age. Encouraged by our previous results in Gran Dolina, we carried out a study to extend and deepen our rockmagnetic investigation of the paleoenvironmental reconstruction of the upper Gran Dolina cave based on experiments that include composition, relative concentration, and grain size of the magnetic iron oxides present in the sediments. Based on the rockmagnetic experiments, we identified magnetite, hematite, goethite, and possibly maghemite in changeable amounts along the profile, which allows us to complement the existing shortage in the literature on the palaeoenvironmental reconstruction of the site. We tentatively interpret the rockmagnetic changes recorded in the cave sediments in terms of glacial/interglacial conditions, furnishing the base for a better understanding for the formation conditions of this unprecedented archaeological site.

1. Introduction

Environmental processes and climatic conditions can be extracted from magnetic minerals that are present in sediments ([1,2,3] and reference therein). With the available modern magnetic instruments, it is possible to detect the presence of trace minerals throughout their magnetic properties, using the dependence to magnetic fields (coercivity) and temperature (Curie point). The key in such paleoclimatic and paleoenvironmental reconstructions is based on three main aspects: composition, concentration, and granulometry of the magnetic minerals.

Caves provide ideal environments for sediment deposition as well as for human activity to furnish protection. Added to UNESCO’s World Heritage List in 2000, the caves of the Sierra de Atapuerca site provide unique testimony of the presence and evolution of the earliest hominids in Europe. The so-called Atapuerca archaeological site includes, among others, the Trinchera del Ferrocarril (Railway Trench), which provides a sign of continuous human occupation from the Early and Middle Pleistocene.

Previous studies based on palynomorphs and faunal remains from the Sierra de Atapuerca caves have attempted to decipher the climatic conditions that were favourable to human presence [4,5,6], but there is still a lot of uncertainty.

For this reason, we carried out a first environmental reconstruction of the upper sedimentary succession of Gran Dolina cave [7] based on the magnetic properties of the sediments. Therefore, encouraged by our initial results, we extended the study to the uppermost stratigraphic units in Gran Dolina using a similar approach. In this paper, after a description of the regional setting and the basics of the archaeological site, we describe the magnetic parameters that we use in order to characterize the composition, concentration, and grain size of the magnetic minerals present in the cave sediments and then finally present a coherent interpretation of the environmental conditions and possible correlation to the Marine Isotope Stage (MIS).

2. Regional Setting and Archaeological Sequence

The Sierra de Atapuerca is a gentle mountain range located in the north-central Iberian Peninsula, within the Duero Basin and between the Ubierna and Demanda mountains, to the north and south, respectively. From a geological point of view, it is a NNW-SSE overturned anticlinal ridge mostly made of Late Cretaceous carbonates. During the Quaternary, and in response to the evolution of the local phreatic condition, a multi-level endokarst system developed in these Cretaceous limestones [8]. The archaeological sites of the “Trinchera del Ferrocarril” (railroad trench) were exposed during the construction of a railway trench running southeast to northwest through the Atapuerca Mountain Range at the beginning of the 20th century. Our study is centred in the Gran Dolina cave (Figure 1), which contains one of the most complete Quaternary stratigraphic sequences with human fossils in Spain. The sedimentary infill is 19 m thick and has traditionally been divided in 11 lithostratigraphic units, TD1–TD11 from the base to the top [9,10]. The formation of the karst and the sedimentary infill respond to the progressive incision of the main river, Arlanzón, starting at least around 1.4 Ma [11]. Shortly after the Jaramillo Subchron (1.07–0.99 Ma), the sedimentary input in the cave is mostly due to gravitational-alluvial sediments [12,13,14,15]. The TD10.1 unit represents the last use of the cave as a central campsite before its final collapse [16].

The chronology of the sedimentary infill is due to paleomagnetism, ESR, U-series, and OSL ([17,18,19,20] and references therein). For this study, it is important to highlight the presence of the Matuyama–Brunhes boundary in the upper part of the TD7 unit, and hence, the Lower–Upper Pleistocene limit. Therefore, units TD1 to TD7 belong to the Lower Pleistocene, whereas TD8 to TD11 belong to the Middle Pleistocene. In this paper, we will focus on the entrance facies of the cave—in particular, the interval between the upper part of TD6 and TD10 units. TD6, the oldest stratigraphic unit with hominid fossils in Gran Dolina [21], was deposited during a warm-temperate period, as revealed by the amphibian and squamate reptile assemblages [22,23,24,25]. The passage to the overlying unit TD7 represents an abrupt climatic change as evidenced in our previous study [7]. TD8 is characterized by at least four units that overall show a decrease in grain size [13]. From an environmental point of view, the presence of thermophilous amphibian and squamates indicates relatively humid-temperate conditions and open environment with permanent water streams. A transition from dry open woodlands to a more open and humid environment has been suggested in the TD8.9 unit [26]. The TD9 unit is described as a phosphatic layer, sterile and composed of bat guano [13]. It has a lenticular shape with its maximum thickness of around one meter to the SE. TD10 unit is the last occupation level from Gran Dolina, dated as Marine Isotope Stage (MIS) 11–9 from geochronological studies [17,27,28,29]. This unit is composed of four subunits (TD10.4 to TD10.1, from base to top) dated between 450 ka and 250 ka [27,28,29] where we find a succession of strata, 3 m thick, associated with breccias and fluvial facies. An environmental change has been detected within TD10 unit [30,31], from the relatively closed-in cavity of TD10.4 to open cave of TD10.1, where we have high density of lithic and faunal remains with little evidences of secondary access and modification by carnivores. Another characteristic of this unit is the presence of two bison bone beds: one in the lowest unit and another in the middle of TD10.2. This latter represents a layer with at least 60 individuals in which no significant post-depositional processes have occurred, and the low-taxonomic diversity excluded important changes in terms of paleoclimate in this sequence [31].

3. Material and Methods

We collected 195 samples along a vertical profile in the Gran Dolina cave in plastic bags using non-magnetic instruments. Due to the current state of the archaeological site, we logged and sampled two complementary different profiles: one comprising units TD6–TD8.9 and another about 5 m to the SE for the units TD9–TD10 (as shown in Figure 1b,c). Our rockmagnetic study was carried out in both the Archaeomagnetism Laboratory at the CENIEH (Burgos, Spain) and Paleomagnetism Laboratory at the University Complutense of Madrid (UCM). The following measurements were performed at CENIEH: susceptibility at two frequencies, isothermal and anhysteretic remanent magnetization, and hysteresis loops. The thermomagnetic study was carried out at UCM.

The bulk susceptibility was measured with an Agico MFK1-FA instrument, obtaining in addition the phase angle—a parameter influenced by all the mineral fractions present in the samples [32] and defined as the difference between in-phase and out-of-phase susceptibility [33,34]. The same instrument was used to measure the susceptibility at two different frequencies, ${\chi }_{lf}$ and ${\chi }_{hf}$, 976 and 15.616 Hz, respectively. We then normalized the frequency-dependent susceptibility (as suggested by [35]) as follows:

$${\chi }_{fn}=\frac{{\chi }_{fd}}{\ln {\chi }_{hf}-\ln {\chi }_{lf}}$$

(1)

where ${\chi }_{fd}=100\left({\chi }_{lf}-{\chi }_{hf}\right)/{\chi }_{lf}\left(\%\right)$—as introduced by [36]—${\chi }_{hf}$ and ${\chi }_{lf}$ are the high and low frequencies, respectively.

We measured the hysteresis cycles with a MicroMag 3900 VSM (Vibrating Sample Magnetometer, LakeShore instrument, MIT Lincoln Laboratory, Cambridge, MA, USA), which is capable of reaching a maximum field of 1.3 Tesla, although we use 0.5 T. The samples used for this instrument were prepared in small gel capsules of ~0.10 mL volume.

Remanence (IRM, ARM) was measured with a three-axes SQUID cryogenic magnetometer (SRM, 2G Enterprises, Applied Physics Systems, Silicon Valley, CA, USA).

Additional progressive IRM acquisition curves up to 5 T were obtained by an impulse magnetizer (ASC Scientific, Model IM10-30, Narragansett, RI, USA), which were used to infer the magnetic components present through their magnetic coercivity.

In order to determine the relative concentration of low to high coercivity minerals, we carried out a particular study based on the application of a saturation magnetization (SIRM) of 5 T, with the help of the impulse magnetizer, followed by alternating field demagnetization at 100 mT to remove the contribution of any ferrimagnetic minerals (e.g., magnetite). Next, we set thermal demagnetization at 130 °C to remove the goethite contribution (method described by [37]).

The grain size and mineralogical composition are inferred from several ratios: the parameter S-ratio provides a measure of the relative amount of high-coercivity to low-coercivity remanence and is obtained by dividing the backwards remanence at 0.3 T with the isothermal remanent magnetization at 1 T. In addition, the ARM/IRM_1T and IRM_1T/χ_lf ratios were also used.

The thermomagnetic curves were obtained at the Paleomagnetism Laboratory at UCM by AGICO KLY-4S Kappabridge coupled with the CS-3 Temperature Control Unit, able to measure the low-field susceptibility variation in the range from ambient temperature to 700 °C.

Lastly, we use the traditional Day plot, which compares the M_rs/M_s with H_cr/H_c ratio and provides information about the domain state of the iron oxides, as well as the King plot to determine the grain size on the basis of the slope given by comparing the ARM susceptibility versus the low-field susceptibility.

For a better comprehension of the manuscript, we include in

Table 1. Acronyms and symbols used in the manuscript.

Equations Acronyms and Symbols
Susceptibility at low field	χlf
Frequency-dependent susceptibility normalized	χfn
Frequency-dependent susceptibility	χfd
High frequency	χhf
Low frequency	χlf
Absolute temperature	T
Magnetic Quantities
Volume susceptibility	Κ
Mass susceptibility	χ
Magnetization	M
Magnetizing field	H
Magnetic induction	B
Saturation magnetization	Ms
Saturation remanence magnetization	Mrs
Coercive force	Hc
Coercivity of remanence	Hcr
Remanent Magnetization
Isothermal remanent magnetization	IRM
Anhysteretic remanent magnetization	ARM
Magnetic Grain Size
Superparamagnetic particles	SP
Single-domain particles	SD
Pseudo-single domain particles	PSD
Multi-domain particles	MD

all acronyms used, and in

Table 2. Parameters and specific plot with their interpretations used in our study (information taken by [1]).

Concentration
Magnetic susceptibility	Estimate of ferrimagnetic material
Frequency-dependent susceptibility	Used to indicate the presence of SP material
Saturation Isothermal Remanent Magnetization	Parameter sensitive to the concentration of principally ferrimagnetic material
Anhysteretic Remanent Magnetization	Measure of the concentration of ferrimagnetic material
Composition
S-ratio	Estimate of the magnetic mineralogy
Thermomagnetic curve	Furnishes the Curie point
Magnetic Grain Size
Frequency-dependent susceptibility	Mostly grain size dependent
King plot	Slope is a function of grain size

we include the used parameters-plot with their interpretation.

4. Results

4.1. Composition Analysis

Initial information of the composition of the studied sediments is furnished by the thermomagnetic evolution of the susceptibility (Figure 2): the red line indicates the heating, and the dashed blue line indicates the cooling curves. We notice in the heating curve that all samples show a susceptibility decrease at 580 °C, which is the Curie temperature of magnetite, and a convex shape upward, typical of ferrimagnetic contribution. In addition, in all the studied units, we can distinguish a weak decrease near the maximum temperature of the experiment (700 °C), which could suggest the presence of hematite [38]. Additionally, the formation of maghemite mineral from magnetite [38] is evident from the increase in magnetic susceptibility up to 250 °C in our heating experiments, although this will not be discussed in this study.

A second slight decrease of the magnetization at around 400 °C is observed in other samples, which is interpretable as the conversion of maghemite in hematite [38] or as moderate substitution degree of magnetite. The cooling presents a susceptibility increase between 580 °C and 450 °C, suggesting a secondary formation of magnetite during the experiment [39] and another progressive increase between 300 °C and 200 °C in all units except TD6.

In order to obtain information on the magnetic components in the sediments through their coercivity, we analysed the IRM progressive acquisition up to 5 T. The resulting curves are normalized by the maximum value (Figure 3). Few samples reveal a rather initial increase (e.g., TD6, TD8.9), suggesting that magnetite dominates the ferromagnetic fraction. Other samples (e.g., TD8) are dominated by a high coercivity component (Figure 3). Sample TD9 possibly shows the highest relative contribution of high coercivity (goethite and hematite).

The different components of the coercivity were analysed using the “Kruiver method” [40], which characterises an assemblage of grains by its SIRM, the median destructive field (B_1/2), and the width of the distribution, identified with Dispersion Parameter (DP) (example for TD9 unit in Figure 4). Taking into account the limits of these parameters (explained by [41]), we infer that in the quasi totality of units there are two main components (a third one could be present, but its contribution is minimal because inferior to 5% of the total IRM), except in the TD8.9 unit where three components are clearly distinguished—as shown in

Table 3. Input parameters (SIRM, B_1/2 and DP) of the components individuated for each unit in the Kruiver approach.

Sample	Component 1				Component 2				Component 3
	SIRM (mA/m)	B1/2 (mT)	DP (mT)	IRM (%)	SIRM (mA/m)	B1/2 (mT)	DP (mT)	IRM (%)	SIRM (mA/m)	B1/2 (mT)	DP (mT)
10.3-546	1.3 × 10−3	338.84	1.8	65	7.1 × 10−4	1258.93	1.8	35
10.3-511	7.8 × 10−4	218.78	2.0	55	6.5 × 10−4	1258.93	2.0	45
10.2-477	8.4 × 10−4	316.23	2.0	53	7.6 × 10−4	794.33	2.0	48
9-586	7.7 × 10−1	794.33	2.1	82	1.7 × 10−1	3162.28	1.3	18
9-569	1 × 10−3	281.84	2.1	61	6.5 × 10−4	1122.02	2.0	39
8.9-492	5.5 × 10−4	251.19	1.8	55	3.7 × 10−4	1412.54	1.7	37	8.5 × 10−5	1584.9	1.9
8.9-483	5.3 × 10−4	165.96	1.9	37	5 × 10−4	416.87	1.9	35	3.9 × 10−4	794.33	2.0
8-522	5.2 × 10−4	338.84	1.9	53	4.7 × 10−4	1412.54	1.9	47

. One feature that deserves to be highlighted is the highest median destructive field found in the second component of a sample taken in the middle-upper part of the TD9 unit (evidenced in light grey in Table 3), which indicates the presence of goethite. As evidenced by previous studies [40,41,42,43,44,45], the discrimination of high coercivity minerals is difficult and depends on their relative percentage on the samples.

To characterize the ferro- and diamagnetic relative contribution, we carried out a number of hysteresis loops (Figure 5). The presence of both para- and ferromagnetism is evident from the shape of the curves. In particular, the loops of samples from units TD9 and TD7 (Figure 5b,e) are a little bit more open than the others, indicating the contribution of high-coercivity minerals such as hematite and goethite [46].

4.2. Concentration Analysis

Low field, bulk magnetic susceptibility is a measure of the concentration of magnetic minerals in the sediments. Low values of K are generally indicative of paramagnetic, antiferromagnetic (hematite), and diamagnetic minerals. Bulk susceptibility (Figure 6a) shows the lowest values in units TD7 and in the upper part of TD9. On the contrary, we see a large ferrimagnetic contribution in the top of TD8.9 and in the subunits of TD10. The phase angle parameter of lower frequency (Figure 6b) presents a slight, yet progressive, increase upward in the profile, only interrupted by minimum values in TD7 and in the middle TD9 units. Finally, the frequency-dependent parameter (Figure 6c) is quite constant with depth but sharply decreases in unit TD9 and lightly in TD7. A possible climatic influence of these changes will be later discussed.

The intensity of the magnetization is given by (a) the saturation magnetization, (b) the anhysteretic remanence (ARM), and (c) isothermal remanence (IRM) (Figure 7). These parameters evidence the great abundance of low coercivity phase, likely magnetite, in TD10 and in the upper part of the TD8.9 unit. A pronounced maximum of ferrimagnetic material is observed in the boundary between TD10.3 and TD10.2, while the upper subunit presents a slight content decrease. The lower values of magnetization are observed in TD7 and in the middle part of TD9, in particular as revealed by the ARM. The overall pattern confirms again different rockmagnetic features of these units, which will be discussed later.

Mineralogy composition and grain size can also be evaluated by some additional ratios (Figure 8). The S-ratio (Figure 8a), defined as IRM₀._3T/IRM_1T, is still the most sensitive parameter for estimating the predominance of antiferromagnetic mineralogy—values near to zero—or ferromagnetic minerals—values near the unity. The entire profile is dominated by magnetite, except in the middle part of the TD9 unit where the presence of a hematite/goethite (high coercivity) mixture is evident. It is important to recall, though, that the presence of high coercivity component could be revealed only if it represents at least 95% of the magnetic fraction, otherwise the magnetite covers it up [47]. The ARM/IRM_1T ratio (Figure 8b) presents a decrease in TD7 and in the middle TD9, while it seems quite constant through the rest of the profile. A key ratio for magnetic grain size is the IRM_1T/χ_lf (Figure 8c), as high values correspond to small size particles, whereas low values correspond to large grain size. The profile variations reveal maximum values in the middle part of unit TD9, indicating the minimum particle grain size. We will contrast this distribution with the King plot.

In order to discriminate the relative concentration of each magnetic component, we carried out a particular study based on the contrasting coercivity values and Curie points. We initially produced a saturation magnetization (SIRM) at 5 T followed by alternating field demagnetization at 100 mT to remove the contribution of any “soft” ferrimagnetic minerals (e.g., magnetite). Next, we performed one step of thermal demagnetization at 130 °C to remove the goethite contribution. Therefore, after these two steps, hematite should be the only magnetic mineral contributing to the total magnetization [37]. We plotted the relative contribution of magnetite versus goethite and hematite, called M/GH (Figure 9a), and goethite versus hematite, G/H (Figure 9b), obtained, respectively, by:

$$\frac{\mathrm{M}}{\mathrm{GH}}=\frac{{\mathrm{SIRM}}_{5\mathrm{T}}-{\mathrm{SIRM}}_{100\mathrm{mT}}}{{\mathrm{SIRM}}_{5\mathrm{T}}}$$

(2)

and

$$\frac{\mathrm{G}}{\mathrm{H}}=\frac{{\mathrm{SIRM}}_{100\mathrm{mT}}-{\mathrm{SIRM}}_{130 °\mathrm{C}}}{{\mathrm{SIRM}}_{100\mathrm{mT}}}$$

(3)

The M/GH ratio (Figure 9a) shows a slight decrease of the magnetite concentration in TD7 unit. In TD8.9, there is an increase in this mineral, whereas in the middle part of TD9, we find the minimum values of magnetite, as also revealed from previous parameters. Finally, in all subunits of TD10, this content is more or less constant. On the other hand, the G/H ratio variation (Figure 9b) is not conclusive at all. Although, a lower goethite concentration in the TD8.9 unit and a higher abundance in the middle-upper part of TD9 are appreciable.

4.3. Grain Size Analysis

Since the magnetite represents the dominant magnetic mineral in the entire profile, we can consider the Day plot [48] in the modified version proposed by [49] as appropriate in order to analyse the magnetic grain size. We have excluded the samples that do not reach the saturation in hysteresis loops for this analysis, as recommended in [50]. All samples analysed fall in the Pseudo-Single Domain (PSD) field (Figure 10).

To further characterize the grain size variation of the particles, we use the so-called King plot [51] that compares the anhysteretic susceptibility (calculated dividing the ARM for the bias field) and the low-field susceptibility; the slope of the alignment defines the grain size of particles. We distinguish three different grain size groups. The smallest grain size family (light blue in Figure 11) corresponds to the samples from the middle-upper part of TD9, coherently with the interpretation of the IRM_1T/χ_lf profile (Figure 8c). A second group is due to the specimens from unit TD7 (yellow in Figure 11), representing the largest magnetic grain size. Finally, most specimens of the other units show an intermediate size between these two extremes.

5. Discussion

The overall TD4–TD11 stratigraphic record in the cave of Gran Dolina is thought to correspond to open-land, with alternation of wet and dry, habitats [6]; however, palaeoenvironmental studies, based on different proxies, confirm vegetation and faunal changes during the interval of the profile that we have analysed (as summarized in

Table 4. Climatic and palaeoenvironmental reconstruction derived by previous studies using different climatic proxies.

Unit	Pollen [4]	Non-Pollen Palynomorphs [5]	Small Vertebrates [6]	Bats [52]	Fish [53]
TD10	Dry and cold conifers woodland	Oscillations of moisture in typical Mediterranean summer	Dry and cold conditions	Low diversity woodland	Decrease in the open-dry habitats and in open humid and woodland ones
TD9		Faeces and other decaying organic material
TD8.9		Early traces of later aridity phase	Open-dry areas with sparse forests	Medium diversity of open woodland
TD8	Warm, humid and wooded conditions	High humidity and decay wood or vegetative remains	Colder climate	High diversity open woodland
TD7	Woodland		Humid and forested areas
TD6.1	Temperate climate	Decrease in local humidity		Water-edge habitat	Top of the coldest period of the entire sequence

). The first pollen studies conducted in this cave [4] evidence a passage from temperate climate of the upper subunit of TD6 to a mostly woodland of TD7–TD8 and especially dry and cold conditions in unit TD10. A successive study on non-pollen palynomorphs [5] interprets the TD8 unit as a decay wood environment with high humidity. These proxies document that the TD10 unit reveals variations of moisture in a typical Mediterranean summer context [5]. Taking into account the small vertebrates [6], TD7 is considered a humid and forested stage passing to open-dry environment with sparse forests in TD8.9 and dry-cold condition in TD10. Additionally, the bat remains [52] confirm a climatic deterioration with an aridity increase and cold pulses that started at the TD8.9 unit and was especially strong at the top of TD10. A recent study, based on fish record of Gran Dolina [53], corroborates the presence of a river system characterized by permanent oxygen-rich, relatively cold (ranging from 0 to 20 °C) running waters. From this point of view, the TD5 unit possibly represents the coldest interval in the entire sequence, while the TD10 unit shows a slight decrease in the open-dry habitats and a minor increase in open humid and woodland ones.

Environmental rockmagnetism has largely been studied in the Chinese loess, showing that, in general, higher values of magnetic susceptibility (the so-called “magnetic enhancement”) correspond to warm conditions which would lead to enhancing pedogenesis processes, which, in turn, promote the production of magnetic minerals, and in particular magnetite (e.g., [1] and references therein). Moreover, a recent study proposed the use of paleosol magnetic properties as proxies of paleorainfall and obtained a correspondence between the pedogenic magnetic susceptibility and mean annual precipitation (MAP) values [54].

Taking into account this information on the environmental reconstruction, we pass to interpret our results. Hematite and goethite usually are the result of oxidative weathering and soil formation [55] with some differences: hematite is favoured by higher soil temperatures and lower moisture content, whereas goethite is promoted by moisture and more acid contexts [32]. Furthermore, the water regime is one of the most important influences on the magnetic properties of a soil: in arid conditions, the pedogenic magnetite could not form, while the water logging could dissolve the magnetic minerals [56]. The presence of hematite over goethite suggests environments with low water activity, such as dry or saline soils or sediments [57].

From our study, we infer a relative decrease in magnetite concentration in respect to the high coercivity components in TD7 based on the S-ratio and bulk susceptibility profile (Figure 6a). Based on the conventional wisdom, we could tentatively correlate level TD7 to a warm time interval. The results of the TD8 unit suggest a distinct magnetite content and a constant high coercivity concentration; moreover, the presence of a stalagmitic crust on the top suggests a warm to temperate climate [17], and it was tentatively attributed to MIS13 (Marine Isotope Stage; [58]). The higher concentration of magnetite and the slight decrease of high coercivity content found in TD8.9 unit can be considered a passage from a humid and cold period. Higher up in the section, the middle-upper part of TD9 unit stands out both because of its visible stratigraphic characteristics (Figure 12) and for its distinctive rockmagnetic properties as shown in this study. As far as stratigraphy, the upper part of TD9 constitutes a rather thick unit of finer grain sediments, in contrast to the underlying and overlying units made up by coarse breccias. Additionally, the lamination and the colour changes configure this level as a distinctive interval in the whole section.

Resuming the results, the middle-upper part of TD9 unit presents the following: low magnetic susceptibility (Figure 6a) compared to the entire sequence; a decrease of the superparamagnetic contribution (as evidenced by frequency-dependent susceptibility in Figure 6c); and especially an abundance of hard ferromagnetic minerals (goethite and hematite) as shown by lower values of S-ratio (Figure 8a) and by IRM progressive acquisition curve (light blue in Figure 3), far from saturation. Moreover, such level presents the smallest magnetic particles of the entire profile, as detected by maximum values of IRM_1T/χ_lf (Figure 8c) and by the King plot (Figure 11). Previous studies also conducted also cave environments [59,60,61] reveal that the increase in susceptibility, correlated to the presence of magnetite, could be due to warm climatic periods, possibly for the pedogenesis processes mentioned earlier. Therefore, the decrease of susceptibility observed in this middle-upper part of the TD9 unit could suggest a priori, the opposite, a cold interval inhibiting the formation and/or deposition of magnetite.

An additional prominent feature distinguishes the TD9 unit, namely the presence of goethite as inferred from the coercivity study that shows extreme values of B_1/2 (more than 3000 mT, as summarized in Table 3). The presence of goethite could cover up the SP particles present in this level and could indicate a rather humid environment for the conditions necessary to form this mineral.

The age dating results could help our proposed environmental reconstruction. There exist a number of age data for the upper levels of the Gran Dolina cave. In particular, a first study [13] using ESR and U-series on teeth assigns to TD8 a mean age of 602 ± 52 ka and 372 ± 33 ka for TD10. For this last, the authors suggested a rapid sedimentation rate and a possible correlation to MIS11. Based on thermoluminescence (TL) and infrared-stimulated-luminescence (IRSL) methods [28], subsequent ages were obtained: 960 ± 120 ka for TD7; 820 ± 140 ka for TD8; 480 ± 130 ka for TD9; and a time interval between 430 ± 59 and 240 ± 44 ka for TD10, although, later on, [62] discussed the validity of such TL results and furnished new ages based on thermally transferred optically stimulated-luminescence (TT-OSL), namely 856 ± 75 ka for TD6, which is in excellent agreement with new paleomagnetic data [14]. In [63], a new age of 418 ± 113 ka was added for the TD9 unit. Finally, ESR data on optically bleached quartz grains extracted from sediments (ESR-OB, [29]) furnished new values. All of these results are reported in

Table 5. Datation of upper sedimentary infill of the Gran Dolina cave obtained by different methods.

Unit	ESR/U-Series (ka) [17]	TL and IRSL (ka) [28]	U-Series TIMS (ka) [63]	ESR-OB (ka) [29]	TT-OSL (ka) [62]
TD10	372 ± 33	244 ± 26 430 ± 59		301 ± 40 458 ± 39
TD9		480 ± 130	418 ± 113 427 ± 267	544 ± 94 565 ± 56 566 ± 81
TD8	602 ± 52	820 ± 140		520 ± 129 473 ± 26 525 ± 26
TD7		960 ± 140		734 ± 128 852 ± 144
TD6					856 ± 75

.

The bulk of these numerical ages should be taken into account when attempting to correlate the stratigraphic units to the corresponding Marine Isotope Stage (MIS) scale. A recently published study [64] furnishes ages of 425 ka and 360 ka to MIS11, which is in agreement with the attribution of the TD10 unit to this stage (as suggested by [17]). MIS11 is defined as a period of reduced global ice volume with numerous interstadial-stadial events [65] and composed of five substages: the 11a, 11c, and 11e relatively warm intervals and the relatively cold MIS11b and 11d [64]. Such alternating warm–cold intervals make it possible that the middle-upper part of TD9, which we think represents a cold interval with high moisture conditions with the concomitant goethite relative abundance, corresponds to MIS11.

6. Conclusions

The rock-magnetism analysis allows us to discern three different magnetic components present in the cave sediments in the upper infill of Gran Dolina, namely units TD7 to TD10: magnetite, goethite, and hematite. The relative concentration variations along the profile are thought to reflect climatic changes during the sediment deposition in the cave. In particular, we suggest a shift from warm to cold climatic conditions from units TD6 to TD7, characterised by a decrease in magnetic susceptibility, a lower concentration of magnetite including SP particles. A similar pattern is observed in the middle-upper part of unit TD9, where we noticed among the lowest values of bulk susceptibility—decrease of magnetite—and the presence of goethite. These observations suggest a cold environment with relatively high moisture in order to account for the production of high coercivity component. Last, we tentatively assign units TD9 and TD10 to the MIS11 stage (also suggested in [13,23,24,25]), which agrees with previous interpretation of the stratigraphic record [27,53]. Ongoing research on both chronology and rockmagnetism will help further interpret the fossiliferous and archaeological units in terms of the climatic shifts during the Quaternary.