Tracing the influence of Mediterranean climate on Southeastern Europe during the past 350,000 years

Loess-palaeosol sequences are valuable archives of past environmental changes. Although regional palaeoclimatic trends and conditions in Southeastern Europe have been inferred from loess sequences, large scale forcing mechanisms responsible for their formation have yet to be determined. Southeastern Europe is a climatically sensitive region, existing under the strong influence of both Mediterranean and continental climates. Establishment of the spatial and temporal evolution and interaction of these climatic areas is essential to understand the mechanisms of loess formation. Here we present high-resolution grain-size, environmental magnetic, spectrophotometric and geochemical data from the Stalać section in the Central Balkans (Serbia) for the past ~350,000 years. The goal of this study is to determine the influence of the Mediterranean climate during this period. Data show that the Central Balkans were under different atmospheric circulation regimes, especially during Marine Isotope Stages 9 and 7, while continental climate prevailed further north. We observe a general weakening of the Mediterranean climate influence with time. Our data suggest that Marine Isotope Stage 5 was the first interglacial in the Central Balkans that had continental climate characteristics. This prominent shift in climatic conditions resulted in unexpectedly warm and humid conditions during the last glacial.


Stratigraphy of the studied profiles
vertic characteristics is exposed (S2), although it is less strongly developed than the palaeosol below. Typical porous loess (L2) is exposed this soil, from 3.45 to 4.4 m. This loess layer is in turn overlain by a very weak, grey soil horizon with a thickness of ca. 0.25 m (covering S1 -L1SS1). From 4.65 to 5 m, the uppermost and youngest loess layer is found (L1LL1). Within the top of this loess (from 5.0 to 5.55 m) the modern soil (S0) is exposed, with a 0.1 m thick Ah horizon and with granular structure at the bottom, an Ah (mollic) horizon with granular structure in the middle and an Ap horizon on top (strongly affected by human activities showing debris of bricks).

Composite profile splicing
Major parts of the five separately sampled profiles were spliced to obtain a composite profile from the Stalać section. The stratigraphy of profiles 2 to 5 was evident from the inclined loess and palaeosol units on one brickyard terrace; sampling was performed in order to have some overlap between these sub-profiles. Supplementary Fig. 4 shows the overlap of some physical and chemical property data used for the splice (magnetic susceptibility, U-ratio, L*, a*, CaO and Cl). Sampling the inclined profiles 2-5 may potentially entail the sedimentary succession with its real -vertical -sediment accumulation. When assuming non-vertical accumulation, but accumulation perpendicular to the slopes, the sampling depth is overestimating the actual sedimentation thickness, and might somewhat mix non-syndepositional sediment. Due to the high sampling resolution and the clear autocorrelation of the data we do not regard this as an issue, but it is important for the understanding of the sedimentary succession.
Profile 1 covers the past ~350,000 year. However, the past 191 ka (S2-S0) is preserved in very low resolution and possibly in discontinuous sedimentation and/or preservation. Therefore, we used only the lower part of profile 1 from upper part of L4 to end of S2 for the splicing. Profile 2 covers the period of formation of the upper part of S2 and the lower part of L2. Since this profile does not cover the whole S2, it is very complicated to splice the upper part of S2 palaeosol (S2SS1) from profile 1 to whole S2 palaeosol from profile 1 because they exhibit different temporal resolutions. Thus, the profile was spliced at the end of S2 at profile 1 and on the beginning of L2 at profile 2. In general, the correlation of profiles 1 and 2 is not straight forward since those profiles formed in different geomorphological settings. It has to be stressed that profile 2 to 5 were deposited on the slope and thus may be formed under different depositional regimes. Thus, the transition from profile 1 to profile 2 was treated with special attention while interpreting the data. Despite differences in some of the bulk data (e.g. colour), we argue that general palaeoclimate signals and trends can be clearly extracted, since similar patterns are observed on the L2 layers of both profiles. However, it has to be noted that it is not possible to directly compare the sedimentation rates from profile 1 to profiles 2-5. Therefore, this study does not focus on the sedimentation rates, especially because sedimentation rates may not be representative for sediments deposited on the slope.
Splicing of profiles 2 and 3 was straight forward. We connected the profiles via the tephra layer ( Supplementary Fig. 4).
Splicing of profiles 3 and 4 was a bit more complicated. The palaeosol pedocomplexes developed on the slope may have overprinted the sediment below in a different way.
However, it seems that this effect is negligible, since most of the proxies show very similar values on both profiles just at the transition from loess to the S1 palaeosol ( Supplementary Fig. 4).

Chronology and age model
Geochronology of loess-palaeosol sequences in the Middle Danube Basin has generally been established mainly by 14 C and luminescence dating for the younger time periods [2][3][4][5] .
Nevertheless, many loess-palaeosol sequences spanning several glacial cycles cannot be dated by these methods. Some sections were dated by correlation of the magnetic susceptibility (χ) signal to oxygen isotope records of marine benthic foraminifera or orbital variations [6][7][8][9][10][11] . Such correlative age models may not be accurate on the scale of few thousand years, but in the absence of other dating techniques this method is generally accepted to provide reliable timescales. Since the χ at the Stalać section is strongly biased by provenance change, we applied correlation of odd Marine Isotope Stages (MIS) to phases of soil formation. The first chronostratigraphy of Stalać based on correlation of the soils to interglacials and loess to glacials was established more than a decade ago 12 . However, our investigations including tephra correlations suggest that this may be more challenging and that the previous chronology is not correct. This fundamentally changes our understanding of this region, suggesting that the previously One of approaches for obtaining age models used in Southeastern Europe is orbital tuning 6,9 . Basarin et al. 6 orbitally tuned a loess record in the Middle Danube Basin. Their record spans almost the last million years, which allows for the determination of specific spectral (frequency) properties in depth and time. Our record spans ca. 350 ka, which is at the very low end of a useful length for orbital tuning. This is especially the case as the sedimentological patterns seem to be dominated by global and regional climate changes on glacial-interglacial level, and do not dominantly correspond to one of the Milankovitch frequencies (precession, obliquity, eccentricity). Therefore we refrain from a tuning approach here, but use a correlation to other proxy records to establish an age model instead.
Two tephra layers were identified in studied profiles at the Stalać section. The lower tephra likely corresponds to the L2 tephra observed in many loess-palaeosol sequences in the Middle Danube Basin 14,17,18 . Although this tephra layer is clearly visible in the field, it underwent significant alteration and glass shards are not preserved. The upper tephra was not visible in the field but the analyzed data clearly pointed to a tephra layer   (2)

Central Balkan
The grain-size records generally show remarkably coarser grain distributions  Table 1 presents the χ of sediments from the possible source areas. We argue that at Stalać changes in χ are mainly determined by changes in source area rather than by pedogenic processes, and can therefore not be considered as a reliable palaeoclimate proxy. Also changing wind intensity 40 Table 1). In general, the grain-size analysis is a well-accepted method in inferring palaeoclimate forcing upon the formation of loess-palaeosol sequences 45,46 , where domination of coarse grain-size particles is usually associated with relatively strong winds and cold climate, while the domination of fine particles is associated with less strong wind, enhanced chemical weathering and warmer climate. Thus, the grain-size data from the following glacial (MIS 8, or loess L3) suggest a rather dry and cold environment ( Supplementary Fig. 5). High values of Ni and Cr ( Supplementary Fig. 11 Fig. 7) and presence of volcanic glass shards ( Supplementary Fig. 8). This clearly indicates a volcanic ash layer that we relate to the Campi Flegrei eruption at 39 ka that has produced widespread tephra deposits throughout the Balkans, lower Danube and northern Pontic area 20,21 (Supplementary  Table 6 where it is shown to which layer, MIS, and height on single and composite profile (if presented) the samples correspond.