Aeolian Dynamics in Southeastern Europe During the Late Pleistocene, Based on Detailed Sedimentological and Geochemical Investigations on Loess

Aeolian dynamics in Southeastern Europe during the Late Pleistocene, based on detailed sedimentological and geochemical investigations on loess

The Faculty of Georesources and Materials Engineering of the

RWTH Aachen University

Submitted by

Igor Obreht, MSc

from Novi Sad

in respect of the academic degree of

Doctor of Natural Sciences

approved thesis

Advisors:

Prof. Dr. rer. nat. Frank Lehmkuhl

Dr Thomas Stevens

Date of the oral exam: 19. April 2017

This thesis is available online on the website of the RWTH Aachen University library.

Table of contents

List of tables I List of figures IV 1. Introduction 1

1.1. Rationale 1

1.2. Objectives 3

1.3. The CRC 806 project –“Our Way to Europe” 6 2. Methods 8

2.1. Sampling strategy and field protocols 8

2.2. Grain-size analyses 9

2.2.1. Background 9

2.2.2. Methodology 11

2.3. Environmental magnetism analyses 12

2.3.1. Background 12

2.3.2. Methodology 13

2.4. Geochemistry analyses 14

2.4.1. Bulk sediment geochemistry analysis 14

2.4.2. Glass shard chemical analysis 16

2.5. Spectrophotometric analyses 16

2.6. Luminescence dating 17 3. Study area and regional settings 20

3.1. Geology and tectonics 21

3.2. Geomorphology 25

3.3. Climate regime 27

3.4. Hydrology 30

3.5. Pedology 33 4. Loess: Rock, sediment or soil - What is missing for its definition? 34

4.1. Introduction 34

4.2. Significance of loess and related deposits 37

4.3. Loess definitions through time 38

4.3.1. Leonhard (1824) and Lyell (1834): a soil-like sediment of alluvial origin 38 4.3.2. Richthofen (1882): aeolian origin of silt particles 39

4.3.3. Berg, 1916 and Berg, 1964): in situ formed soil (regardless of parent material) 39

4.3.4. Ložek (1965): complex formation of loess environment 39

4.3.5. Liu (1988): more than a deposit of dust storms 40

4.3.6. Pye (1987 and 1995): windblown silt deposit 40

4.3.7. Pécsi and Richter, (1996) and Pécsi (1995): not just the accumulation of dust 41

4.3.8. Recent developments 42

4.4. Loess genesis and the role of loessification 45

4.4.1. What is loessification? 45

4.4.2. The relevance of carbonate and clay for loess formation 46

4.4.3. Sedimentary processes: the prerequisite for loessification? 47

4.4.4. Particle trapping: the onset of loessification? 48

4.4.5. The problem of polygenetic loesses 49

4.4.6. Diagenesis 50

4.4.7. Pedogenesis 50

4.5. Dust, loess, and loess-like sediments 51

4.6. Conclusions 53 5. Aeolian dynamics at the Orlovat loess–palaeosol sequence, northern Serbia, based on detailed textural and geochemical evidence 54

5.1. Introduction 55

5.2. The Orlovat site in northeastern Serbia 57

5.2.1. Study site and geomorphological setting 57

5.2.2. Sampling strategy 57

5.3. Material and methods 59

5.3.1. Grain-size and geochemical analysis 59

5.3.1.1. Ratios and chemical weathering indices 60

5.4. Results 60

5.4.1. Grain-size distributions and their change with stratigraphy 60

5.4.2. Geochemical analyses 64

5.4.3. Stratigraphy and chronology 68

5.5. Discussion 70

5.5.1. Granulometry as an indicator of wind dynamics 70 5.5.2. A grain-size perspective for the reconstruction of palaeoenvironmental conditions 73

5.5.3. Geochemical characteristics (simple ratios and weathering indices) 75

5.5.4. Potential changes in source material 76

5.6. Conclusions 79 6. Tracing the influence of Mediterranean climate on Southeastern Europe during the past 350,000 years 81

6.1. Introduction 82

6.2. Results 85

6.2.1. Stratigraphy and chronology 85

6.2.2. Particle size properties and environmental magnetism 86

6.2.3. Bulk sediment geochemistry 87

6.2.4. Tephrochronology 88

6.2.5. Spectrophotometric results 88

6.3. Discussion 88

6.4. Methods 97 7. Shift of large-scale atmospheric systems over Europe during late MIS 3 and implications for Modern Human dispersal 100

7.1. Introduction 101

7.2. Regional settings and study sites 103

7.3. Results 104

7.3.1. Luminescence dating 109

7.4. Discussion 109

7.5. Material and methods 116 8. Synthesis 118 9. Abstract 122 10. Zusammenfassung 125 11. Acknowledgments and author contributions 128 12. References 132 Supplementary Material Chapter 5 168 Supplementary Material Chapter 6 174 Supplementary Material Chapter 7 203

List of Tables

Page

Table 5.1. 69

Correlation of lithological units with Marine isotope stages (MIS) and depth.

Supplemantary Tables:

Supplementary Table 5.1 169

Detailed major (wt%) and trace element (ppm) composition of the Orlovat section.

Supplementary Table 6.1. 175

Ni, Cr and magnetic susceptibility values from the Zapadna, Južna and Velika Morava rivers alluvium samples

Supplementary Table 6.2. 186

Tuning points of Stalać age model to Marine Isotope Stages based on LR04 stack (Lisiecki and Raymo, 2005)

Supplementary Table 6.3. 187

Major oxide geochemical results from microprobe analyses of glass shards from upper crypto tephra layer (the Campanian Ignimbrite/Y-5) at the Stalać section. Data are presented as raw values. Analytical settings used for determining the glass-shard major oxides composition is presented in the Supplementary Table 6.4.

Supplementary Table 6.4. 189

Analytical settings used for determining the glass-shard major oxides composition at Bayerisches GeoInstitut, Bayreuth University. Order of measuring elements (first to last): Na, Si, K, Ca, Fe, Mg, Al, P, Ti, Mn, Cl-.

Supplementary Table 6.5. 193

Summary of the De, dose rate and resulting pIRIR age data (Bösken et al., 2017). Water content was obtained in the laboratory. Equivalent doses (De) are shown as result of central

age model (Galbraith et al., 1999). Standard errors are indicated. Ages are expressed with a 1- sigma error range. Supplementary Table 6.6. 200

Codes related to the samples presented in Supplementary Fig. 6.10. The first column contains the code, the second presents the layer and MIS to which the samples correspond, the third presents the height of the samples on the original profile and the fourth column presents the height of the samples on the composite profile (if presented)

Supplementary Table 7.1. 209

Summary of luminescence data for samples from Urluia (URL1). IRSL measurements were done on polymineralic fine grains (4-11 µm) on 9.8 mm aliquots using the pIRIR290 protocol (Buylaert et al., 2012; Thiel et al., 2011). An internal K content of 12.5 ± 0.5 % (Huntley and Baril, 1997) and Rb content of 400 ± 100 ppm (Huntley and Hancock, 2001) was assumed. The cosmic dose rates were calculated according to Prescott and Hutton (1994), taking into consideration the altitude (125 m), latitude (44.094167°N) and longitude (27.9031°E) of the sampling site, as well as the density (1.7 g cm-3) and thickness of the overlying sediments.

Supplementary Table 7.2. 212

Major oxide geochemical results from microprobe analyses of glass shards of the tephra layer (Campanian Ignimbrite/Y-5) at the Vlasca section. Data are presented as raw values. Analytical settings used for determining the glass-shard major oxides composition is presented in Table S7.3.

Supplementary Table 7.3. 213

Supplementary Table 7.4. 221

Tuning points of the Urluia and Vlasca sections correlation to δ18O from Greenland ice sheet (Andersen et al., 2004).

Supplementary Table 7.5. 226

Tie points of the final age models for the Urluia and Vlasca sections.

III

List of Figures

Page

Figure 3.1 21

Study area: map of Southeastern Europe. Map is showing loess sequences studied in this dissertation (yellow circles; Orlovat, Stalać, Urluia and Vlasca) and other key loess sequences (yellow rectangles; Dunaszekco, Ruma, Titel, Batajnica, Belotinac, Rasova) and lacustrine records (red rectangles; Ohrid, Prespa and Tenaghi Philippon) discussed in the dissertation.

Figure 3.2. 22

Geological units of the Alpine-Carpathian-Dinaridic orogenic systems. Red dots present the studied sections (modified after Dimitrijević, 2002).

Figure 3.3. 24

Sketch of southern tip of the Tornquist–Teisseyre Line (modified after Hippolyte, 2002). NDO = North Dobrogean Orogen; CD = Central Dobrogea; SD = South Dobrogea; FB = Carpathian Foreland Basin; HCM = Holy Cross Mts; PCF = Peceneaga–Camena Fault; RRF = Rava– Ruska Fault; SIF = Siret Fault; SOF = Solca Fault.

Figure 3.4. 29

Climate diagrams of Zrenjanin, Kruševac (Serbia), and Constanta (Romania). The diagrams were created on the basis of the climatological normal for the period 1991-2010 for Zrenjanin and Stalać (dotted lines represent a time period from 1961-1990), and from 1961–1990 for Constanta (WMO (World Meteorological Organization), 1996). Periods of drought occur when the mean monthly precipitation curve is below the mean monthly temperatures curve on the diagram (Walter, 1974).

Figure 4.1. 36

Loess(-like) deposits. A: Haarlass, Heidelberg (Central Europe). Type section of loess (Leonhard, 1824). Photo by C. Hornung (Heidelberg). B: Batajnica, Serbia (SE-Europe). One of few plateau-like loess deposits in Europe with quasi-continuous loess-palaesol sequences. C: Landscape in the surrounding of Luochuan, Chinese loess plateau. Largest loess deposit in the world with longest LPS and specific loess morphology. Photo by Z. Svirčev (Novi Sad), IV

D: Caspian Lowlands, Iran, close to the border to Turkmenistan: Small mesas (10-20 cm height) in the semi-desert composed mainly of silty to fine sandy sediments. Mean annual precipitation: ~ 100-200 mm. The darker mesas are connected to cyanobacterial mats which partly prevent the silty parent material from erosion.

Figure 4.2. 42

The Central part of the European loess map of Haase et al. (2007) with abbreviated capital names. Localities mentioned in Figs. 4.1 and 4.3 and political boundaries of Hungary indicated. Note the rather exclusive appearance of alluvial loess in this country's area.

Figure 4.3. 44

Polygenetic loess in the Krems region, Lower Austria, possibly comparable to derasion loess of Pécsi and Richter (1996). A: Morphological situation of Krems Schießstätte (white rectangle). Location at E-exposed slope (i.e. leeward position to W-winds) on a spur between Danube and Krems River. B: View of the outcrop Schießstätte: Mostly a wall-stable yellowish substrate dominated by silt, i.e. a loess-like sediment. C: Polygenetic loess from the northern part of the outcrop. Note the distinct gravel bands (partly discontinuous) and the presence of gravel throughout the sequences indicated by small holes due to profile cleaning (some marked). D: Aggregates of loess-like sediments from this profile. Upper one weakly structured, lower one with slight biogenic structure (size of aggregates: 4 x 2.5 cm).

Figure 5.1. 58

The study area. (A) Map of the Vojvodina region with the geographical positions of the main loess sections (Markovićet al., 2014a, modified). (B) A geomorphological map of the Tamiš loess plateau surrounded by the Tamiš and Begej river valleys (Popov et al., 2012, modified).

Figure 5.2. 61

Density distribution curve for the main stratigraphic units of the Orlovat section.

Figure 5.3. 61

The grain-size proxies, U-ratio, GSI and CaCO3 content related to the pedostratigraphy. Ages shown in ka next to the sequence represent the results of luminescence dating (Marković et al., 2014a).

Figure 5.4. 63

Direct comparison between >44 μm fractions obtained from the Orlovat, Titel (Marković et al., 2008) and Surduk (Antoine et al., 2009). The profiles are plotted on their depth scales.

Figure 5.5. 65

The rock magnetic proxies (Marković et al., 2014a), major elements (normalised and presented in percentages values) and CaCO3 content related to the pedostratigraphy.

Figure 5.6. 66

The rock magnetic proxies (Marković et al., 2014a), trace elements (presented in ppm values) and CaCO3 content related to the pedostratigraphy.

Figure 5.7. 68

The rock magnetic proxies (Marković et al., 2014a) simple ratios, weathering indices and related CaCO3 content related to the pedostratigraphy. Ages shown in ka next to the sequence represent the results of luminescence dating (Marković et al., 2014a).

Figure 5.8. 78

A–CN–K diagram.

Figure 6.1. 84

Map of the Balkan Peninsula, Middle and Lower Danube Basins, showing key loess-palaeosol sequences (yellow rectangles; Stalać (this study), Ruma (Vandenberghe et al., 2014), Titel (Bokhorst et al., 2011), Batajnica (Buggle et al., 2014, 2013, 2009), Orlovat (Marković et al., 2014a; Obreht et al., 2015), Belotinac (Basarin et al., 2011; Obreht et al., 2014), Mircea Voda (Buggle et al., 2014, 2013, 2009)) and lacustrine records (red rectangles; Ohrid (Sadori et al., 2016), Prespa (Leng et al., 2013) and Tenaghi Philippon (Tzedakis et al., 2006)) discussed in this paper. The white dashed line represents the current northern limit of the Mediterranean climate. The map was generated using ArcGIS 10.2.2 (http://www.esri.com/software/arcgis/arcgis-for-desktop/free-trial).

Figure 6.2. 86

Clay fractions, U-ratio, χ, Ni contribution, L* and a* values (see Methods) related to pedostratigraphy of the composite profile from the Stalać section.

Figure 6.3. 90

Direct comparison between the benthic δ18O LR04 stack (Lisiecki and Raymo, 2005), MEDSTACK planktic δ18O data (Wang et al., 2010), arboreal pollen from Tenaghi Philippon (Tzedakis et al., 2006), arboreal pollen (Pinus, Juniperus and Betula are excluded) from the Lake Ohrid core (Sadori et al., 2016), U-ratio and <2 µm fractions from the Stalać section (plotted versus age, abscissa), U-ratio and <2 µm fractions from the Ruma section (Vandenberghe et al., 2014) (the only section with existing grain-size record spanning the last three glacial-interglacial cycle in Middle Danube Basin; plotted vs. depth). Note the differences between scales on the plots presenting Stalać and Ruma grain-size data. Tephra layers are marked with yellow lines; “CI” refers to the Campanian Ignimbrite/Y-5 tephra layer and “L2” refers to the L2 tephra layer.

Figure 6.4. 92

Progressive termination of Mediterranean influence over Southeastern Europe, as indicated by the succession of palaeosol types and <5µm particles peak values in related palaeosols at (a) the Stalać section (Central Balkans) and (b) the Batajnica-Stari Slankamen spliced section (Middle Danube Basin; Buggle et al., 2013).

Figure 7.1. 103

Map of the Southeastern Europe showing key loess-palaeosol sequences (Urluia (this study), Vlasca (this study), Rasova (Zeeden et al., in press), Titel (Basarin et al., 2014; Bokhorst et al., 2011), Batajnica (Buggle et al., 2014, 2013), Orlovat (Marković et al., 2014; Obreht et al., 2015), Stalać (Obreht et al., 2016), Dunaszekcső (Újvári et al., 2016)) and lacustrine records (Lake Prespa (Panagiotopoulos et al., 2014) and Tenaghi Philippon (Tzedakis et al., 2006)) discussed in this paper. The map was generated using ArcGIS 10.2.2 (http://www.esri.com/software/arcgis).

VII

Figure 7.2. 105

Direct comparison between proxies from the Urluia and Vlasca sections represented by U- ration, fine particles (<5 µm) and χfd (values are normalised), and their comparison with a stacked climatic record from northern China (CHILOMOS (Yang and Ding, 2014)) and δ18O record from Dim Cave (Ünal-İmer et al., 2015) over the past 55,000 years. The straight yellow line represents the timing of the Campanian Ignimbrite tephra deposition. Black arrows indicate a trend of general continentalization.

Figure 7.3. 106

Direct comparison between χfd record from Urluia (green line) and Vlasca (blue line), χfd from Titel loess-palaeosol sequence (Basarin et al., 2014) (orange line; Middle Danube Basin) and their comparison with the mean annual lake surface temperature (LST; red line) of the Black Sea (Wegwerth et al., 2015) over 25,000-50,000 years ago.

Figure 7.4. 108

Weathering indices (CIA and CPA) of the Urluia and Vlasca sections

Figure 7.5. 114 a) Simplified scheme of general atmospheric circulation patterns over Europe during middle MIS 3. Note that the westerlies were reaching Central and Southeastern Europe via the NW- SE trajectory; b) Simplified scheme of general atmospheric circulation patterns over Europe during late MIS 3. Increased Siberian High influence on Europe had a major influence on Eastern and Western Europe, while the westerlies shifted to the S-E trajectory, bringing the warmer air masses from the Mediterranean to the Balkans and the Middle Danube Basin. Blue line represents a schematic intensification of Siberian High, green lines show prevailing air masses and red lines represent paths of AMH dispersal in Europe. The map was generated using ArcGIS 10.2.2 (http://www.esri.com/software/arcgis).

Supplementary Figures:

Supplementary Figure 5.1. 168

The grain-size categories normalised to <63 µm related to the pedostratigraphy

Supplementary Figure 6.1. 175 VIII

Photo of the stratigraphic section sampled (profile 1) during the first field campaign. The profile is marked with a red rectangle. It preserves a record of aeolian sediment over the past ~350,000 years.

Supplementary Figure 6.2. 177

Photos of the stratigraphic sections, with labels to aid in correlation. Photos are of the central profile a) and the south-western part sampled during the second field campaign b) - profile 2- 4 and c) – profile 5. Sampled areas of the profiles are marked with red rectangles. The tephra layer found in profiles 1, 2 and 3 corresponds to the same layer.

Supplementary Figure 6.3. 178

Correlation of the profiles: a) profile 1 (as in Supplementary Fig. 6.2), b) profiles 2-4 (as in Supplementary Fig. 6.2), c) indication of profile 5. Due to semicircular excavation shape of this site, profile 5 is not shown here but on Supplementary Fig. 6.2.

Supplementary Figure 6.4. 181

Splicing of profiles 1-5. Dashed black lines present the splicing points. Light lines of records present the part of profiles that were not used in the composite profile. Note that the upper part of profile 1 is compressed, revealing a different sedimentation rate. The difference in pattern between profiles 1 and 2 arises from different sedimentation rates; tephra occurrence provides a certain marker in L2 facilitating correlation between profiles 1 and 2.

Supplementary Figure 6.5. 182

Grain-size distribution of each particle size class and cumulative distribution from all grain- size classes (each colour in the right box represent a corresponding class in µm) of the composite profile.

Supplementary Figure 6.6. 182

Magnetic susceptibility, frequency dependent magnetic susceptibility, L*, a* and b* values related to pedostratigraphy of the composite profile.

Supplementary Figure 6.7. 186

Correlation of height scale and the resulting age assignment of different units at Stalać. Bulk sediment geochemical proxy Cl and magnetic susceptibility respectively, provided useful proxies for the identification of tephra layers. On the left, preliminary luminescence ages (Marković et al., 2006) are presented.

Supplementary Figure 6.8. 191

Microscopical photos of volcanic glass shards from cryptotephra L1SS1LLL1 layer. The images were taken with magnification factor 50. The sediment was suspended in water and viewed in transmitted light.

Supplementary Figure 6.9. 192

Total alkali - Silica diagram (modified after Anechitei-Deacu et al. (2014)) showing the geochemical correlation between the Campanian Ignimbrite tephra identified at Stalać and other regional occurrences, including proximal pyroclastic flow (Civetta et al., 1997) and plinian fall deposits in Italy (Signorelli et al., 1999), as well distal fine ash occurrences within Mediterranean marine records (Pyle et al., 2006) and in the Russian loess (Pyle et al., 2006), the terrestrial sequence at Caciulatesti in southern Romania (Veres et al., 2013), and Urluia (Fitzsimmons et al., 2013) and Rasova – Valea cu Pietre (Anechitei-Deacu et al., 2014) loess/palaeosol sequences in the Lower Danube loess. Inset: scanning electron microscope (SEM) image of the Campanian Ignimbrite glass shards from Stalać.

Supplementary Figure 6.10. 199

Density distribution curves of grain size analyses of representative samples from all stratigraphic units at each sampled profile. Representative samples of one unit from each profile are presented by colour and codes that correspond to the numbers from 1 to 35. Explanation of each sample code is presented in Supplementary Table 6.6 where it is shown to which layer, MIS, and height on single and composite profile (if presented) the samples correspond.

Supplementary Figure 6.11. 201

Ni and Cr concentrations presented on age scale. Green dashed lines present mean value of samples from the Great Morava alluvium sediment, red dashed lines present mean value of X

samples from the Zapadna Morava alluvium sediment and yellow dashed lines present mean value of samples from the Južna Morava alluvium sediment. Values located in the light red rectangle are under relative domination of the Zapadna Morava alluvium as source area, while values in light yellow rectangle are under relative domination of the Južna Morava alluvium as source area.

Supplementary Figure 6.12. 202

Mean grain-size (µm) and L* values from the Stalać section compared to the normalised modelled Greenland ice-sheet volume (de Boer et al., 2014a). Note that the transition from MIS 7 to MIS 6 at the Stalać section (transition from S2 to L2 layer) has to be considered with caution as explained in Supplementary chapter 4.

Supplementary Figure 7.1. 207

Dose recovery test (top) and prior IR stimulation temperature test (bottom). All tested samples are within the desired recovered/given dose ratio of 1.0±0.1. A plateau for prior IR stimulation temperatures is present between 50 and 170°C.

Supplementary Figure 7.2. 208

Consistency between luminescence ages (red squares and uncertainty), correlative age model (grey line; black dots represent tie points) and the final age model (black line) from the Urluia section. Orange lines represent the timing and the depth of the Campanian Ignimbrite tephra layer.

Supplementary Figure 7.3. 211

Scanning electron microscope (SEM) images of the Campanian Ignimbrite glass shards from Vlasca.

Supplementary Figure 7.4. 222

Correlation between the δ18O data from Greenland ice sheet (Andersen et al., 2004) (blue line) and proxies from the Urluia section (green lines for χfd (see the discussion in chapter 4.2 for the explanation of two different scales), purple line for U-ratio, red line for fine particles (<5 µm) and orange line represents mean grain-size ).

Supplementary Figure 7.5. 223

Correlation between the δ18O data from Greenland ice sheet (Andersen et al., 2004) (blue line) and proxies from the Vlasca section (green line for χfd, purple line for U-ratio, red line for fine particles (<5 µm) and orange line for mean grain-size). Grey rectangle represents a sandy layer that is not considered for palaeoclimate reconstruction (not shown elsewhere).

Supplementary Figure 7.6. 224

Correlation between the δ18O data from Greenland ice sheet (Andersen et al., 2004) (blue line), and χ and χfd from Vlasca (orange and red lines), Urluia (lines in different shades of green; see chapter 4.2 for the explanation of two different scales) and Rasova (Zeeden et al., in press) (lines in different shades of purple) sections presented on the correlative age models. Grey rectangles represent Heinrich Events, straight yellow line represent the timing of the Campanian Ignimbrite tephra deposition.

Supplementary Figure 7.7. 225

Comparison of mean grain-size from Vlasca (blue line) and Urluia (green line), and the accumulation rate of coastal ice-rafted detritus (IRDC) from the Black Sea as an indication of winter severity (Wegwerth et al., 2015a, 2015b).

Supplementary Figure 7.8. 227

Comparison of the correlative age model (red lines) and final age model (black lines). Note the same general trends regardless which age model is used.

Supplementary Figure 7.9. 228

Major elements from the Urluia and Vlasca sections. Major elements are presented in %.

Supplementary Figure 7.10. 229

Geochemical ratios commonly used to indicate possible changes in the provenance. No ratio indicates a major change in provenance, only Al2O3/K2O ratio at Vlasca suggests a slight increase in K-feldspars between 20-10 ka.

XII

Supplementary Figure 7.11. 230

Frequency dependent magnetic susceptibility from Rasova (Zeeden et al., in press) (red line), Vlasca (blue line) and Urluia (green line), U-ratio from Vlasca (pale blue line) and Urluia (dark green line) compared to Land Evolution Zones (LEZ) form laminated Eifel maar sediments (Sirocko et al., 2016) of the last 60,000 years. Note the similar timing of the transition of LEZ 7 (boreal forest) to 6 (steppe) and continentalization of the Lower Danube Basin.

XIII

1. Introduction 1.1. Rationale

Knowledge of past climate dynamics is crucial for our understanding of the Earth’s present- day state and climate predictions (e.g. Caseldine et al., 2010; Ganopolski et al., 2016; Yin and Berger, 2015). The basis of our understanding of the past climate (palaeoclimate) is grounded in the sedimentary archives. Marine sediments (e.g. Zachos et al., 2001) and ice cores (e.g. Andersen et al., 2004) are the most representative archives for the reconstruction of the global or hemispherical palaeoclimate. Nonetheless, to reconstruct the regional and supraregional palaeoclimates of the continents, sediments from a lake (lacustrine), and peat bog environments, as well as aeolian sediments are the key archives (Feurdean et al., 2014). Especially in arid and semi-arid areas, aeolian sediments are usually the only available past climatic records. Aeolian sediments originate from particulate material that was transported and accumulated by the wind. Among the aeolian sediments, loess sequences are widely spread, especially in the middle latitude continental parts of the Northern Hemisphere (Pye, 1995; Smalley et al., 2011). Loess is commonly defined as windblown dust, with a high contribution of silt predominantly consisting of silicates and detrital carbonate. However, this definition oversimplifies processes that occur upon loess deposition. The term of loessification is used to describe the postdepositional process of accumulated mineral dust (Pécsi, 1990). Although the mechanism of loessification is still debatable, whether there is more pedogenetic or diagenetic influence, it comprises a process of initial silicate weathering, partial carbonate dissolution and re-precipitation, and neo-formation of clay minerals.

Overall, loess sediments provide sensitive records of past climatic and environmental changes. As loess is formed from silts transported by winds, it sensitively records temporal dynamics of past near-surface wind systems during its deposition. Also, the understanding of postdepositional processes provides an insight into the environmental conditions during and after dust deposition. Loess and loess-related deposits are one of the most widespread Quaternary formations, especially in the continental Eurasia (Smalley et al., 2011). Overall, loess is covering more than 10% of the earth's surface, mainly in the Earth's temperate zone (Pye, 1995; Muhs and Bettis, 2003). Some parts of the Eurasian loess belt has preserved quasi-continuous records of glacial-interglacial changes during the Quaternary, and in exceptional cases, loess formation may have started as early as during the Miocene (Guo et al., 2002). 1

In Europe, the best preserved loess-palaeosol sequences are in Southeastern Europe. In this region, the formation of thick loess sequences that contain a valuable palaeoclimatic record started during the Early Pleistocene (Marković et al., 2011). However, more often, loess- palaeosol sequences span a time interval from the Middle Pleistocene (e.g. Buggle et al., 2013; Fitzsimmons et al., 2012; Marković et al., 2008; Necula et al., 2015; Vandenberghe, 2013; Zeeden et al., 2016) or the Late Pleistocene (Basarin et al., 2011; Marković et al., 2014a; Stevens et al., 2011) until present. Southeastern Europe is a geomorphologically diverse region, mainly mountainous throughout the south (the Dinarides, Rhodope and Balkan Mountains), whereas in the north the Carpathians separate two large lowland basins, the Middle Danube (Carpathian) Basin in the west and the Lower Danube (Walachian) Basin in the east. As a consequence of its geomorphology, Southeastern Europe is under the influence of the Mediterranean, Atlantic and continental climate zones. This makes Southeastern Europe potentially very sensitive to changes in climate and renders it probably the best region for reconstructing the interaction of European large-scale atmospheric systems.

Over the last decade, Southeastern Europe (in particular the Middle and Lower Danube Basins) has been in the focus of the Quaternary research in Europe (Buggle et al., 2008; Fitzsimmons et al., 2012; Marković et al., 2015; Újvári et al., 2008). As a result, our knowledge of the Quaternary climate and environmental conditions in the region has been fundamentally improved. However, those studies did not establish the patterns of the atmospheric circulations responsible for loess formation. For example, the understanding of loess formation at the Chinese Loess Plateau is fundamental for understanding the evolution of the East Asian winter and summer monsoons. The observed patterns in loess-palaeosol sequences can even give information on climate forcing mechanisms unavailable from other geoarchives (Hao et al., 2012, 2015) and improve predictions of the future climate dynamics (Yang et al., 2015). Unfortunately, the driving mechanisms of European loess-palaeosol sequence formation are not yet well understood. The biggest challenge is to distinguish between the more relevant atmospheric circulations for loess formation, and the ones that have played a minor role. However, in addition to the three main atmospheric zones (Mediterranean, Atlantic and continental), there are also several local atmospheric circulations (e.g. the Košava wind in the southeastern part of the Carpathian Basin). Therefore, establishing the development of atmospheric circulations in different parts of Southeast

Europe would be a significant step towards the general understanding of the past intensities and interactions of the local, regional and large-scale atmospheric circulations.

1.2. Objectives

The basic aim of this dissertation is to gain information on the Late Quaternary climate and environmental history inferred from four Southeastern Europe loess sections (Orlovat (Serbia), Stalać (Serbia), Urluia (Romania) and Vlasca (Romania)) based on detailed high- resolution grain-size, environmental magnetism, and geochemical analyses. The main goal is to improve the knowledge of aeolian dynamics over Southeastern Europe mostly during the Late Pleistocene and to provide the first information about the evolution and interactions of Mediterranean, Atlantic and continental climate regimes in this time span. Albeit Southeastern Europe was under the focus of the Quaternary loess research over the past decade, mostly the regional palaeoclimate and palaeoenvironmental conditions were reported and the attention was directed at finding similarities rather than difference from various regions. The objective of this dissertation is to report similarities and differences between the sections in the same region to distinguish local and regional signals, and then to compare different region in Southeaster Europe to reconstruct their climatic evolution and past interaction. However, before any conclusions on palaeoclimate can be inferred from loess sequences, it is crucial to understand the processes of loess formation, to establish a reliable chronology, and to evaluate the potential and time limits of the studied archives and applied proxies. Although the main goal of this dissertation is not to reveal and better understand the dominant sedimentary processes, chronological issues and evaluation of the proxies relevant in loess research, those questions will also be addressed during the evaluation of the data and before making any palaeoclimatic statements.

The dissertation is divided into eight main chapters adhering to the objectives and topics stated above. Chapter 1 is an introducing chapter and presents a general idea and objectives of this dissertation. It is followed by Chapter 2 which exhibits the protocols of the field work and the fundaments of the used methods. Grain-size analyses, studies of environmental magnetism and geochemical element composition present a fundamental approach in the methodology

used for this dissertation. Other methods and techniques that were used are also briefly described.

Chapter 3 introduces the study area. It reviews the geological background and the knowledge of tectonics in the studied region, geomorphological settings, current climate regimes, hydrology and distribution of soils and vegetation.

Chapter 4 provides theoretical knowledge of loess as sediment, but also as a palaeoclimatic archive. Besides the historical overview of loess research throughout history, it presents a comprehensive overview of processes that precede loess accumulation and the formation of loess. Although this chapter does not deal with the study area, this chapter represents one of the foundations on which the following chapters are based on. It gives essential information on loess and a palaeoclimatic archive and gives a basic introduction to principles how it is used in the palaeoclimatic reconstruction. Accordingly, it is grounded in all following chapters. This chapter is published in Quaternary International 399, 208-217 (2016).

Chapter 5 addresses some of the knowledge gaps left by the previous studies at the Orlovat section (Lukić et al., 2014; Marković et al., 2014a). Unusual regional characteristics inferred from the Orlovat loess section were reported using environmental magnetism and colour of the sediment, with a support of luminescence dating. Chapter 5 describes main sedimentological and geochemical characteristics of Orlovat loess in order to give an explanation of the unique characteristics. The aeolian dynamics in South Banat and the influence of the Košava wind are also explained. Further, the chapter gives information on the river dynamics and change in provenance and source area. This chapter is published in Aeolian Research 18, 69-81 (2015).

Chapter 6 establishes the spatial evolution of the Mediterranean-like climate over the Balkans and Southeastern Europe during the past 350,000 years. It is based on the high-resolution grain-size analyses as well as environmental magnetic, spectrophotometric and geochemical data from the Stalać section in the Central Balkans (Serbia) for the past ~350,000 years. The significance of this chapter is because it for the first time establishes the border of the influence of one large-scale atmospheric system (Mediterranean climate in this case) over Southeaster Europe. It also provides valuable information on the last glacial over the Balkans region that would not be possible studying last glacial cycle only. This chapter is published in Scientific Reports 6, 36334 (2016). 4

In Chapter 7, the high-resolution dataset from the Lower Danube Basin is presented. The palaeoclimatic conditions in the Lower Danube Basin over the past ~50,000 years are reconstructed. Since in Chapter 6 the limits of Mediterranean climate influence is established, in this chapter results from the Lower Danube Basin were compared with the Middle Danube Basin to gain a general an idea of the large-scale atmospheric systems mostly between Atlantic and continental climates. By comparing the obtained results with other palaeoclimatic archives from other Eurasian regions, an overall large-scale atmospheric circulation for Europe during the late MIS 3 is proposed. This chapter is published in Scientific Reports 7, 5848 (2017).

A synthesis of the main results and findings is presented in Chapter 8. This chapter puts this dissertation into a general context of palaeoclimate research and suggests possible directions for future studies.

1.3.The CRC 806 project –“Our Way to Europe”

This doctoral dissertation has been carried out within the framework of the second phase of the Collaborative Research Centre (CRC; “Sonderforschungsbereich” or SFB) project 806 called “Our Way to Europe”. This project is running as a collaboration between the University of Cologne, RWTH Aachen University and the University of Bonn. “Our Way to Europe” studies a specific part of the history of mankind using a combination of geoscientific and archaeological methods. “Our Way to Europe” concerns the complex chronological, regional, climatic, environmental and socio-cultural aspects of the major intercontinental and transcontinental events related to the dispersal of Anatomically Modern Humans from Africa to Europe. It covers the time span between the dispersal of Anatomically Modern Humans from Africa (190,000 years ago) to the permanent establishment of Anatomically Modern Humans in Central Europe. This project is designed to concentrate in particular on three major themes:

1) The climatic, environmental and cultural contexts of the primary expansion of our species, particularly Anatomically Modern Humans dispersal from Africa ~190,000 years ago and the occupation of Europe ~40,000 years ago.

2) Secondary phases of expansion and retreat of our species induced by climatic, environmental or cultural changes, for instance, the reoccupation of Near East in the Middle Weichselian and the re-occupation of extensive parts of Europe after the end of the Last Glacial Maximum, resulting in the spread and establishment of the Neolithic economy throughout Europe.

3) Population changes, mobility and migration, driven by the growing impact of human agency on the environment, particularly dispersal, retreat and internal mobility among sedentary prehistoric societies.

Due to the complexity of the spatial and temporal relations pertinent to the project questions, “Our Way to Europe” is divided into seven clusters. Clusters A-D (A: Northeastern Africa; B: Near East, Anatolia and the Balkans; C: Northwestern Africa and Iberian Peninsula; D: Central Europe) are related to specific geographical regions. Cluster E focuses on dating and cluster F on contextual changes, while cluster Z deals with management, administrative and financial issues. The study area of this dissertation was produced within the cluster B, dealing

with the Eastern corridor of Anatomically Modern Human dispersal into Europe (Near East, Anatolia and the Balkans). The presented results are the contribution to the subproject B1. The main focus of subproject B1 (The „Eastern Trajectory“: Last Glacial Palaeogeography and Archaeology of the Eastern Mediterranean and of the Balkan Peninsula) lies on the elucidation of the archaeology and the environmental/ecological conditions of the last glacial, and the relation of climate and humans. The special focus is on the time period ~40,000 years ago, which is believed to be the timing of the appearance of Anatomically Modern Humans in Europe. This dissertation gives a contribution to the understanding of climatic conditions in the studied region over the Last Glacial and considers possible relations between climatic conditions and Anatomically Modern Human dispersal.

2. Methods 2.1. Sampling strategy and field protocols

Selection of the sections was done with a great care, following the idea to compare the data of the sections from different regions in Southeastern Europe and on the various geomorphological positions to better understand between local and regional deposition modes and palaeoclimatic signals.

The Orlovat section (Chapter 5) was chosen because it has already been studied (Lukić et al., 2014; Marković et al., 2014a), and it was indicated by rock magnetic properties, the colour of the sediment and luminescence dating that specific sedimentological and/or environmental conditions dominated. However, previous studies did not succeed to give more information on deposition modes and environmental and climatic evolution on the Orlovat section. To close this gap, high-resolution grain-size analyses were performed to understand in more details the deposition modes and patterns at the site, while the geochemical analyses were performed to obtain a better understanding of the provenance of sediment in the studied region.

The Stalać section (Chapter 6) was selected as a key section in the Balkans since it represents one of the best-preserved sections, with the highest sedimentation rates and the longest time span in the Balkans region. Therefore, by comparing the Stalać section with already studied sections in the other parts of Southeastern Europe, it was expected to gain new information about the palaeoclimate over the Balkans. This study presents a multi-proxy data-set of grain- size, rock magnetics, colour and geochemical analyses of the sediment.

The Urluia and Vlasca sections (Chapter 7) were chosen since they are located in generally poorly investigated part of Southeastern Europe (the Lower Danube Basin), and also because both sections have archived a uniquely high-resolution palaeoclimatic record with unusually high mass accumulation rates for European standards. The idea was to compare adjacent high sedimentation rate sections with different geomorphological characteristics, where Vlasca is located on the Danube River, and Urluia is located ~50 km to the south, and to understand whether those records preserve a local, regional or large scale climatic signal. Before sampling, the topmost 10-20 cm of exposed material was removed from all profiles, allowing for continuous, high-resolution, incremental sampling at all five sections. After the profile cleaning, a detailed profile description was conducted. Sampling was performed over three different profiles from summer 2013 to autumn 2015. The sampling of the Orlovat section 8

(Chapter 5) took place earlier and the laboratory analyses were conducted for this dissertation. The Stalać section (Chapter 6) was sampled in five separate subsections that were clearly correlated with palaeosol or tephra layer. The sampling was conducted in the 5 cm increments (with the exception of 20 samples that were sampled in 2.5 cm increments). For the study presented in Chapter 7, the upper parts of the Urluia and Vlasca sections were sampled in 2 cm increments.

Sampling for luminescence dating was performed on all sections. However, only seven samples from the Urluia section are directly involved in this dissertation. At the Orlovat section (Chater 5) dating was part of previous studies (Marković et al., 2014a; Timar-Gabor et al., 2015). For the Stalać section (Chapter 6) luminescence dating was performed on 6 samples in a separated publication (Bösken et al., 2017), although a general overview of the samples characteristics and preliminary ages are also presented here. At Urluia, seven samples for uppermost 7 m were dated and presented in this dissertation. At Vlasca several luminescence samples were taken for the whole last glacial, however, those will be dated in the future. All samples for luminescence dating were taken by hammering and extracting 15 cm long metal tubes into the sediment wall, and by collecting the surrounding material in a radius of 30 cm for dose rate determination.

2.2. Grain-size analyses 2.2.1. Background

Grain-size measurements are a fundamental method used in this dissertation and it was applied for all original research presented here (Chapters 5, 6 and 7). Commonly, the analysis of grain-size distribution is accepted as a reliable method for determining the characteristics of aeolian transport processes (Újvári et al., 2016; Vandenberghe, 2013). Wind transport of the particles is traditionally divided into three basic categories: rolling, jumping over short distances near the surface (or saltation) and suspension. The rate and distance of such transport strongly depend on the availability of particles, grain properties and wind strength. Accordingly, aeolian dust transport is not a uniform process. The complexity of aeolian transport was reported already half a century ago by e.g. Doeglas (1968) and Folk (1966). In loess research, detailed reviews on the physical background of particle transport and deposition mechanisms were presented by e.g. Pye (1987), Tsoar and Pye (1987) and recently by Újvári et al. (2016). However, numerous grain-size studies based on loess sediment tend to

oversimplify those processes. Although it is challenging to distinguish between processes influencing particles transport, this dissertation is paying a special attention to those problems.

Especially in loess research, additional problems for understanding the transport mechanisms arise from postdepositional weathering of particles and formation of clay. Therefore, it is important to know the main environmental conditions under which loess forms. According to Lehmkuhl (1997), Quaternary loess was mostly accumulated in areas with steppe vegetation, which was constrained to an annual precipitation amount of 250-600 mm. In drier areas vegetation was too sparse to bind loess, while in more wet areas soil development was more extensive and vegetation too dense. Recent studies have suggested that loess may also form under drier conditions due to the presence of biological crusted surfaces and their abilities to catch dust particles (Smalley et al., 2011; Svirčev et al., 2013). These newest findings indicate that loess might be formed under different environmental conditions, additionally complicating the comparison of loess sequences.

The differences in the methodology related to loess analyses also make the comparison between different studies even more challenging. Traditional methods, such as a combination of sieving and pipette analyses, are generally straightforward methods. Unfortunately, they are very time-consuming and the obtained results exhibit a rather low size resolution and limited possibilities for establishing useful grain-size ratios. On the contrary, laser diffraction analyses allow an insight into a wider fraction size range (Pye and Blott, 2004). Those methods are also less time-consuming (Özer et al., 2010). However, the laser diffraction analyses show big limitations in establishing reliable proportions of fine fractions, making the comparability of results by different methods challenging (Konert and Vandenberghe, 1997). The core of the problem lies in the non-spherical shape on the fine particles, especially clay. Fine fractions with non-spherical shape are classified to the larger particles if they are illuminated orthogonally, while it is classified to the smaller size if the particle is parallel to light. Therefore, the laser diffraction analyses lead to an overall underestimation of clay contribution (Konert and Vandenberghe, 1997). Also, different devices for laser diffraction analyses may yield different results.

In the studied area both sieving and pipette analyses (Marković et al., 2006, 2007) and laser diffraction analyses (Antoine et al., 2009; Bokhorst et al., 2011; Necula et al., 2013; Obreht et al., 2014; Vandenberghe et al., 2014) were used. Although laser diffraction analyses were

obtained from different devices, most of the studies classify clay in range of <4 µm up to <5.5 µm (Antoine et al., 2009; Bokhorst et al., 2011; Necula et al., 2013; Obreht et al., 2014; Vandenberghe et al., 2014). The grain-size results used in this dissertation were obtained by a Beckmann‐Coulter LS13320 device. This device uses a technology known as Polarization Intensity Differential of Scattered Light (PIDS) which improves the accuracy of fine particle measurements by illuminating the sample with three additional wavelengths of 450 nm, 600 nm and 900 nm (Buurman et al., 1997). Also, the Mie optical model was applied to partly overcome the problem of non-spherical shape of fine particles (Buurman et al., 1997). Thus, the clay in this dissertation is classified as <2 µm (Schulte et al., 2016). Appling this methodology increased the precision of the fine particles but additionally complicated the comparison with the previous studies. The fine particles are also used to indicate weathering and pedogenesis, and therefore they present an important proxy for palaeoclimate reconstruction. Since the aim of this dissertation is to establish both local and supraregional climatic patterns recorded in loess, the fractions with <5 µm grain size are also considered here to keep coherence with the previous studies.

2.2.2. Methodology

All the samples were air-dried, homogenised and sieved to fractions <2 mm in diameter. Subsamples of 0.1–0.3 g fine-earth (< 2mm) were pre-treated with 0.70 ml of 20% hydrogen peroxide (H2O2) at 70 °C for 12 hours. This process was repeated until a bleaching of the sediment occurred (Allen and Thornley, 2004), but not beyond three days. To keep particles dispersed, the samples were treated with 1.25 ml, 0.1 M sodiumpyrophosphate . (Na4P2O7 10H2O) for 12 h (Pye and Blott, 2004). The particle size was measured by the LS 13320 Laser Diffraction Particle Size Analyser (Beckman Coulter) equipped with Polarization Intensity Differential of Scattered Light (PIDS) technology. This instrument yields volumetric percentage frequencies of 116 grain size classes from 0.04 to 2000 μm with an error of 2% (1σ). Each sample was measured four times at two different concentrations to increase the accuracy. Afterward, all measurements with a reliable obscuration were averaged. To calculate the grain-size distribution the Mie theory was used (Fluid RI: 1.33; Sample RI: 1.55; Imaginary RI: 0.1) (Özer et al., 2010; ISO 13320-1, 1999). The particle size fractions were defined by employing the ISO standard 14688 (2002), where clay is represented with particles smaller than 2 μm, fine silt from 2 to 6.3 μm, medium silt from 6.3 to 20 μm and coarse silt from 20 to 63 μm (Blott and Pye, 2012). 11

2.3. Environmental magnetism analyses 2.3.1. Background

Magnetic susceptibility and other rock magnetics properties are commonly used proxies in the loess research. In this dissertation, low-frequency magnetic susceptibility and frequency dependent magnetic susceptibility were used in Chapters 5-7 as important proxies for palaeoclimate reconstruction (albeit for Chapter 5 it is originally presented in Marković et al. (2014a). Low-frequency magnetic susceptibility is often used as a proxy for increased weathering/pedogenesis and soil/sediment moisture (Buggle et al., 2009, 2014; Hao et al., 2008; Marković et al., 2009; Necula et al., 2013). The principles behind this proxy in loess are based on a control of magnetic susceptibility which is determined by the amount and composition of iron-bearing paramagnetic and ferromagnetic minerals and their grain-size distribution. Especially important are magnetite and maghemite (ferromagnetic minerals), since they have several magnitudes higher magnetic susceptibility than hematite and goethite (antiferromagnetic minerals) (Buggle et al., 2009; Thompson and Oldfield, 1986). Since ferromagnetic minerals are predominantly formed during pedogenesis, even small amounts can significantly increase magnetic susceptibility and give an indication on palaeoenvironment.

Another important factor controlling magnetic susceptibility is the grain-size distribution magnetic particles. Magnetic grains of superparamagnetic size particles (<~30 nm) have a significantly higher magnetic susceptibility than stable single-domain particles (>~30 nm particles with only one region with parallel coupled atomic magnetic moments) or multidomain particles (>~10 µm particles with several regions with parallel coupled atomic magnetic moments) (Thompson and Oldfield, 1986). In major parts of Eurasia where the loess is formed on the plateau (area such as the Chinese Loess Plateau or loess plateaus in the Lower and Middle Danube Basins), it has been demonstrated that superparamagnetic size particles are predominant in palaeosols because those particles precipitate from weathering solutions in the processes occurring during soil formation (e.g. Buggle et al., 2009; Evans and Heller, 2003; Heller et al., 1991; Heller and Liu, 1984; Maher, 2011). Therefore, in these regions magnetic susceptibility represents a reliable proxy for pedogenesis. The most important factors for the formation of ferromagnetics are soil temperature, soil moisture, pH value and content of organic matter (Evans and Heller, 2001). Nonetheless, a reduction of magnetic susceptibility in palaeosols, when compared with intercalated loess units, has been 12

reported from high latitude Alaskan and Siberian loess deposits. This phenomenon has been explained by an increased wind strength during glacial periods, which more efficiently transports dense iron oxide particles (e.g. Evans, 2001). Soils were developed during warmer intervals when aeolian transport of dense magnetic particles from the parent sources did not play an important role. However, such processes are generally negligible in middle latitude Eurasian loess belt (Buggle et al., 2014; Zeeden et al., in press).

Although magnetic susceptibility is a widely used proxy in loess research concerning the middle latitude Eurasian loess belt, this proxy is slightly biased to the grain-size distribution and change in the source area because multidomain particles are still important contributors of low-field magnetic susceptibility enhancement. However, multidomain can be detrital, and therefore more influenced by a change in the source area. Moreover, magnetic susceptibility is also influenced by the change in the concentration of low-magnetic carbonate content. Using frequency dependence magnetic susceptibility in great extent overcomes the above stated flaws (albeit there are also several other ways to get a better insight into the pure pedogenic signals, see e.g. Liu et al. (2014)). Frequency dependence magnetic susceptibility refers to the difference of low to high-frequency susceptibility values (here calculate as χfd = (χlf – χhf) / χlf

*100 [%], where χlf and χhf are low-frequency and high-frequency susceptibility) and therefore it is not influenced by carbonate content. But more important, frequency dependence magnetic susceptibility is determined by superparamagnetic and single domain grain size particles that are mostly products of pedogenesis and therefore much less influenced by the change in the grain-size distribution during deposition and changes in source areas. Although frequency dependent magnetic susceptibility does not fully rule out possible detrital signal, it provides a highly sensitive proxy mainly for humidity during dust accumulation (Buggle et al., 2014; Zeeden et al., in press). The frequency dependence magnetic susceptibility presents a measure for the relative contribution of superparamagnetic-ferrimagnetic particles close to the superparamagnetic–stable single-domain particles threshold, and therefore it can very sensitively record for even small changes in pedogenesis intensity.

2.3.2. Methodology

Environmental magnetism analyses for the Stalać section in Chapter 6 and the Vlasca section in Chapter 7 (the same methodology was used for results presented in the Chapter 5 (the Orlovat section), but previously presented in Marković et al. (2014a)) were carried out on

bulk samples. The dried sediment was filled into 6 cm3 plastic boxes, and subsequently compressed and fixed with cotton wool to prevent movement of sediment particles during measurement. The volumetric magnetic susceptibility was measured at frequencies of 300 and 3000 Hz in a static field of 300 mA/m using a Magnon International VSFM. Data were corrected for drift and for the effect of sampling boxes (weak diamagnetism), and normalised to density. Hence, magnetic susceptibility is given as mass specific susceptibility in m3/kg. For the Chapter 7, the upper part of the Urluia section samples were treated in the same way. However, 323 samples from the Urluia section were sampled as oriented samples, taken using brass tubes and an orientation holder. The samples were placed in the diamagnetic boxes directly in the field. Beside sample preparation, further measurements were the same. The frequency dependence was calculated as χfd = (χlf – χhf) / χlf *100 [%], where χlf represents low field magnetic susceptibility and χhf represents high field magnetic susceptibility (Buggle et al., 2014; Heller et al., 1991; Liu et al., 2014; Worm, 1998).

2.4.Geochemistry analyses 2.4.1. Bulk sediment geochemistry analysis 2.4.1.1. Background

Geochemistry analysis of loess constitutes a powerful tool for indicating the provenance, source area and weathering of sediment. In this dissertation, elemental geochemistry was conducted in Chapters 5-7. Especially in Chapter 7, geochemistry is used to establish the intensity of the weathering by chemical weathering indices. Generally, chemical weathering indices are based on the concept of mineral alteration, where the selective removal of soluble and mobile elements from a profile section is compared to a relative enrichment of immobile and non-soluble elements (e.g. Buggle et al., 2011; Fedo et al., 1995; Harnois, 1988; Kronberg and Nesbitt, 1981; Nesbitt and Young, 1982; Yang et al., 2004). Numerous simple ratios of bulk elements and ratios of several elements have been used for the reconstruction of palaeoenvironmental conditions of loess sequences (Bokhorst et al., 2009; Buggle et al., 2011; Kels et al., 2014; Muhs et al., 2008; Varga et al., 2011). However, a robust understanding of the weathering inferred from those ratios is still missing. During the last few years, some studies attempted to establish the most appropriate weathering ratio that is independent of other proxies for loess sequences, such as grain-size or source area/provenance change (Buggle et al., 2011; Schatz et al., 2015; Yang et al., 2006). An extensive overview of the evaluation of weathering indices is presented by Buggle et al. (2011). 14

The most important criteria for weathering indices are establishing the most appropriate soluble and mobile element, and also choosing the most appropriate immobile non-soluble element. The elements as Al, Si, Ti, Ka, Ba, Rb and Zr are most frequently used for non- soluble elements in weathering indices because those elements form insoluble hydrolyzates (Buggle et al., 2011). However, Ti, Zr, and Si are relatively sensitive to changes in parent material composition, while K, Ba, Rb are not always recommended since under intense weathering conditions significant losses of these elements can occur during the transformation of micas, feldspars and other host minerals into secondary clay minerals (Gallet et al., 1996; Muhset al., 2001). For mobile and soluble elements Ca, Mg and Sr can be used since they are common in silicate minerals such as plagioclase, pyroxene, amphibole and biotite, which are susceptible to weathering (Nesbitt et al., 1980). However, in a parent material containing carbonate the mobility of these elements is predominantly controlled by the behaviour of calcite and dolomite, wich make these elements not very useful for weathering indices in loess. Probably the best selection of mobile elements is on Na and K since those elements are hosted in the same mineral group (feldspars) as Al. Therefore, the widely used Chemical

Index of Alteration (CIA) = Al2O3/(Al2O3+Na2O+CaO*+K2O)) * 100 (where CaO* is silicatic CaO; Nesbitt and Young, 1982) is probably one of the most appropriate weathering indices. However, Buggle et al. (2011) specified ratios relying only on Na as the only soluble element, and established the Chemical Proxy of Alteration (CPA) = (Al2O3/(Al2O3+Na2O)) * 100 (also known as CIW´ (Cullers, 2000)) as the best geochemical proxy of silicate weathering for loess-palaeosol sequences, in order to exclude potential effects arising fromthe inconsistency in K weathering. Nevertheless, the ratio between Al2O3, Na2O+CaO* and K2O, usually presented as A-CN-K diagram (Nesbitt and Young, 1984), represents an important ratio informing about weathering and sorting effects of aluminosilicates, as well as the initial composition of the unweathered material (e.g. Nesbitt and Young, 1989; Nesbitt et al., 1996).

An even more important role of geochemical studies could be establishing the provenance or the source area of the sediment material (Buggle et al., 2008; Nie et al., 2015; Újvári et al., 2008, 2012). However, detailed analyses of elemental geochemistry in Southeastern Europe showed that the geochemical results cannot prove an accurate provenance, but rather can rule out some areas as source area (Buggle et al., 2008; Újvári et al., 2008). Despite the limitations in establishing the provenance of the particles, elemental geochemistry has turned out useful

in establishing the source area, confirming previously a proposed source area of the Danube River and its confluence valleys (Smalley and Leach, 1978).

2.4.1.2. Methodology

The bulk sediment samples were sieved down to 63 μm and dried at 105˚C for 12 h. An 8 g- quantity of the sieved material was mixed with 2 g Fluxana Cereox wax, homogenised and pressed to a pellet with a pressure of 20 t for 120 s. The measurements were conducted by means of a pre-calibrated method. Loess and palaeosol samples were analysed in duplicate for major and trace element abundances with polarisation energy dispersive X-ray fluorescence

(EDPXRF) using a SpectroXepos. The CaCO3 content (Chapter 5) was determined volumetrically with the SCHEIBLER-method (Schaller, 2000; ISO 10693, 1995).

2.4.2. Glass shard chemical analysis

In this dissertation, microprobe analyses were used to obtain a geochemical composition of the glass shards found at the Stalać section and to get information of the relevant tephra layer and the age of the layer. The sediment sample was sieved and glass shards were isolated through density separation, mounted in epoxy resin, ground and polished in preparation for microprobe analysis. Measurements were made using single-grain, wavelength-dispersive electron microprobe analysis at the Bayerisches GeoInstitut on a Jeol JXA8200 microprobe employing an accelerating voltage of 15 keV. A 6 nA beam current and a defocused beam were used. The order of measured elements was as follows (first to last): Na, Si, K, Ca, Fe, Mg, Al, P, Ti, Mn, Cl. The peak counting times were 10 s for Na, 30 s for Si, Al, K, Ca, Fe and Mg, 40 s for Ti and Mn, and 60 s for P. Precision was estimated at <1–6% (2σ) and 10– 25% (2σ) for major and minor element concentrations, respectively.

2.5. Spectrophotometric analyses

Spectrophotometric analyses were used at the Orlovat section in Chapter 6. Colour variations as a proxy in loess research have been used for instance as the proxy for variations in mineral concentrations (e.g. Ji et al., 2002; Lukić et al., 2014; Torrent et al., 2007). The colour of the sediment can also be used to get an idea of the pedogenesis intensity and soil formation. A

Konica Minolta CM-5 spectrophotometer was used to determine the colour of dried and homogenised sediment samples by detecting the diffused reflected light under standardised observation conditions (2° Standard Observer, Illuminant C). Spectra were obtained in the visible range (360 to 740 nm), in 10 nm increments, and these data were then converted into the Munsell colour system and the CIELAB Colour Space (L*a*b*) using the Konica Minolta SpectraMagic NX software. The resultant values indicated the extinction of light on a scale from L* 0 (absolute black) to L* 100 (absolute white), and express colour as chromaticity coordinates on red-green (a*) and blue-yellow (b*) scales.

2.6. Luminescence dating

Optically stimulated luminescence (OSL) dating belongs to the techniques of trapped charge dating exploiting the property of certain minerals, such as quartz and feldspar, to act as natural dosimeters (e.g. Ankjærgaard, 2010). These minerals are subject to low levels of radiation by naturally occurring radioactive isotopes of elements like U, Th, Rb and K, where radiation leads to ionisation of the minerals’ atoms (Walker, 2005). Through time, the minerals recorded the amount of radiation they have been exposed to by storing an increasing amount of energy in the form of trapped electrons (Duller, 2015). These electrons can be released in the laboratory by light or heat leading to a recombination of the trapped charge in luminescence centers and emission of light that is proportional to a number of electrons trapped (e.g. Walker, 2005).

The luminescence signal of feldspar saturates later than the one of quartz, which is why measurements on feldspar and polymineral samples are suitable to date older sediments (sediments with higher palaeodoses). The conventionally applied IRSL (infrared stimulated luminescence; e.g. Preusser et al., 2003; Vasiliniuc et al., 2013; Watanuki and Tsukamoto, 2001) of feldspar/polyminerals suffered from anomalous fading, a decrease of the IRSL signal with time faster than expected from thermal stability measurements. A consequence is a signal loss that leads to age underestimation (Spooner, 1994; Wintle, 1973). This is circumvented by applying two infrared stimulations during the pIRIR (post-IR IRSL) measurements; the first stimulation empties unstable electron traps that are thought to be responsible for the signal loss, while only the second (more stable) luminescence signal is used for age calculation. Especially, the pIRIR290 (pIRIR stimulation at 290°C) shows very 17

small fading rates that are often negligible (Buylaert et al., 2012; Thiel et al., 2011; Vasiliniuc et al., 2012).Even though pIRIR protocols (e.g. Buylaert et al., 2012; Thiel et al., 2011; Thomsen et al., 2008) are not as established as the SAR protocol (Murray and Wintle, 2000, 2003), they are nowadays often in environments with high dose rates such as loess. In many sections there has been good agreement between quartz and feldspar dating, particularly for equivalent doses < 200Gy (e.g. Stevens et al., 2011).

Luminescence ages were used in a study presented in Chapter 7. Luminescence analyses were performed in the frame of a PhD project closely related to this dissertation. Moreover, luminescence dates presented in Chapter 6 are originally presented in Bösken et al. (2017). Since luminescence dating is not a focus of this dissertation, no in-depth description of the method and the state of the art will be discussed here. However, the methodology used in this dissertation is presented.

To obtain the ages pIRIR290 protocol was used. Samples for De determination were prepared under subdued red illumination. When opening the steel cylinders the first ~2 cm of sediment from both ends were removed. After drying the samples for 24 h at 50°C the gravimetric water content was calculated after weighing the samples and measuring the loss of water. The dried samples were then treated with 10% hydrochloric acid, 10% hydrogen peroxide and 0.01 N sodium oxalate to remove carbonates and organic matter, and to disperse sediment aggregates. The 4-11 µm grain size fraction was isolated using information from Stokes’ law and by removing the clay fraction (< 4 µm) via centrifuging.

Continuous wave optically stimulated luminescence measurements (CW-OSL) were carried out on a Risø TL/OSL DA 20 reader. The reader was equipped with a 90Sr/90Y β-source and infrared (IR) LEDs emitting at 870 nm (FWHM = 40 nm). A 410 nm interference filter was used for signal detection. The post infrared infrared stimulated luminescence protocol (pIRIR) was used to determine the equivalent dose of polymineral fine grains. (Buylaert et al., 2012; Thiel et al., 2011). The signal was integrated using the first 2.4 s of the stimulation curve minus a background derived from the last 25.6 s. A prior IR stimulation temperature test (Buylaert et al., 2012) was performed on sample C-L3713. Furthermore, several dose recovery tests (DRTs; tests used to ensure that the results are reproducible) with the favourable prior IR stimulation temperature were performed using bleached samples (24 h solar simulator). Finally, a mean equivalent dose was determined using the central age model

(CAM) and the common age model (COM) (Galbraith et al., 1999). Residual doses were assessed after bleaching the aliquots for 24 hours in a Hönle Sol2 solar simulator. Fading tests were conducted on sample C-L3711. For these, a laboratory beta dose of 186 Gy was used and the normalised luminescence signals were measured after different storage times between 100 and 800 minutes.

Samples for dose rate determination were oven-dried at 50 °C for at least 48 h, homogenised and packed into plastic cylinders for radionuclide measurements. Radionuclide concentrations were measured on a high-purity germanium gamma-ray spectrometer after a resting period of four weeks to compensate for radon emanation during pretreatment. The dose rates and ages were calculated using the DRAC programme, v.1.2 (Durcan et al., 2015). Conversion factors of Liritzis et al. (2013) and an estimated water content of 10±5% were included. The internal β-dose rate contribution was calculated by assuming a K content of 12.5±0.5% (Huntley and Baril, 1997) and a Rb content of 400±100 ppm (Huntley and Hancock, 2001). Attenuation factors of Bell (1980) and Guerin et al. (2012) were used. The cosmic dose rate was calculated following Prescott and Hutton (1994) considering the geographical position, altitude, sample depth and density of the overlying sediments. The α-efficiency was determined for six aliquots of two samples. Aliquots were thermally annealed (480 °C) before administering α-doses of up to 200 Gy in a Freiberg Instruments Lexsyg Research device. The laboratory dose of the samples was determined using β-irradiation in a Risø reader (pIRIR protocol). Finally, the α-efficiency was calculated as the ratio of the measured β-dose divided by the given α-dose (Kreutzer et al., 2014). A linear function was used for fitting. A mean a- value was used for all samples.

3. Study area and regional settings

Southeastern Europe (Fig 3.1) is a term that is commonly used for the region wider than the traditional Balkans. Since the term Balkans is used to define a geographical area, but also accepted as a cultural area, the borders of this area are very various and disputed. However, the Balkans, or the Balkan Peninsula, is a well-defined geographical area and in a geomorphological sense, it comprises the peninsula surrounded by the Black Sea to the east, the Mediterranean Sea (including the Ionian and Aegean Seas) and the Marmara Sea to the south, and the Adriatic Sea to the west. Its northern boundary is often regarded as the Sava River and the Danube River after Sava’s confluence into it. Due to the political situation and borders of the Balkans countries, the term Southeastern Europe is used to cover countries that fully or partly spread over the Balkan Peninsula. Accordingly, Southeastern Europe covers Albania, Bosnia and Herzegovina, Bulgaria, Croatia, Greece, Macedonia, Montenegro, Romania, Serbia, Slovenia and the European part of Turkey (in political divisions Greece is often related to Southern Europe and Slovenia is often related to Central Europe). Therefore, in this dissertation, we refer to Southeastern Europe from the geomorphological perspective, where Southeastern Europe covers the Balkan Peninsula, the South Carpathian Mountains, the southern part of the Middle Danube (Carpathian, Pannonian) Basin and the Lower Danube Basin.

Figure 3.1. Study area: map of Southeastern Europe. Map is showing loess sequences studied in this dissertation (yellow circles; Orlovat, Stalać, Urluia and Vlasca) and other key loess sequences (yellow squeres; Dunaszekco, Ruma, Titel, Batajnica, Belotinac, Rasova) and lacustrine records (red squeres; Ohrid, Prespa and Tenaghi Philippon) discussed in the dissertation.

3.1. Geology and tectonics

Tectonics movements in Southeastern Europe were determining precondition for loess formation. Subsidence of the Danube basins enabled the formation of large river networks that are the most important source areas in Southeastern Europe (Buggle et al., 2008). The uplift of the surrounding mountains (Alpes, Carpathian and Dinarides; Figs. 3.1. and 3.2) gave rise to enhanced physical erosion, leading to increased production of the dust particles. Therefore, special attention of this dissertation is paid to the geology and tectonics of the regions where the studies sections are situated. The Orlovat section (Chapter 5) is placed in the south of the Middle Danube Basin (Figs. 3.1 and 3.2), close to the foothills of the Carpathians. The Stalać section (Chapter 6) is on the border of the Rhodopes and Dinarides (Figs. 3.1, 3.2), while the Urluia and Vlasca sections (Chapter 7) are in the Dobrogea region in the Lower Danube Basin (Fig. 3.1, 3.2, 3.3). 21

Figure 3.2. Geological units of the Alpine-Carpathian-Dinaridic orogenic systems. Red dots present the studied sections (modified after Dimitrijević, 2002).

The Orlovat section is located in the Pannonian Basin (also the Carpathian or Middle Danube Basin), a large lowland basin mostly spreading over Central Europe, while the southernmost parts belong to Southeastern Europe (Figs. 3.1 and 3.2). During the Miocene and Pliocene epochs this area was under the Pannonian Sea (Harzhauser and Piller, 2007). The Pannonian Sea was part of the Paratethys Sea, where a Miocene uplift of the Carpathian Mountains isolated the sea from the rest of Paratethys. The Pannonian Sea existed for about 9 million years (e.g. Ivanov et al., 2011). Eventually, the sea lost its connection to Paratethys and became a lake (the Pannonian Lake). Its last remnant, the Slavonian Lake, dried up in the Pleistocene epoch (e.g. Leever et al. 2011). The Quaternary landscape evolution of the Middle Danube Basin was strongly controlled by tectonic processes. The basin inversion with NW– SE and N–S compression was the dominant endogenic triggering mechanism, causing an uplift of the mountains at the basin margin and areas of uplift as well as accelerated

subsidence. This region is largely covered with Quaternary sediments, out of which loess comprises a significant part.

The Stalać section is in the Central Balkans, an area characterised with a very complex geological background (Figs. 3.2). It is built up from three large tectonics units: the Carpatho- Balkan Mountains in the east, the Serbian-Macedonian mass (an extension from the Rhodopes) in the central part, and the inner Dinarides in the west (Fig. 3.2). The boundaries between these three tectonics units are two deep continental rifts with meridional directions (Burchfiel et al., 2008). They can be followed from the Pannonia Basin (the Middle Danube Basin) in the north to the Aegean Sea in the south (Fig. 3.2). Besides these two major rifts, there are many smaller transversal rifts. During the Neogene, the Serbian-Macedonian mass experienced subsidence, while the Dinarides and the Carpatho-Balkan Mountains were uplifted. Therefore, over the Neogene, the Serbian-Macedonian mass was covered with the Pannonian Sea, and mostly Miocene sediments were deposited. After the withdrawal of the Pannonian Sea, the valleys of the Južna and Velika Morava Rivers were formed in this region.

Over the Pliocene and the Quaternary, volcanic activity increased over the continental rifts. Therefore, this area is characterised with volcanic rocks, mostly andesite, rhyolite and trachyte. Recent volcanic activities are indicated by thermal waters in this region. Besides the volcanic rocks, this area is characterised by crystalline schists, Mesozoic limestones and dolomites. In the basins and valleys Neogene lake sediments (clays, sands and conglomerates), as well as Quaternary fluvial (sands and gravels) and aeolian (loess) sediments, are present.

The Urluia and Vlasca sections are situated in the Dobrogea region in the Lower Danube Basin (Fig. 3.1). It is a key region for understanding the geodynamics of Central Europe as it is located in the foreland of the southeastern Carpathians, at the margin of the Black Sea, and at the southeastern extremity of the Tornquist–Teisseyre Line (Fig. 3.3), which connects the Baltic Sea to the Black Sea (Hippolyte, 2002). Dobrogea is the only place south of Poland where the Tornquist–Teisseyre Line, Europe’s longest lineament, outcrops (Hippolyte, 2002). The Dobrogea area comprises two important faults (Peceneaga-Camena and Capidava- Ovidiu) which divide this region into North, Central and South Dobrogea (Fig. 3.3; Posea, 2002). The Northern Dobrogea is characterised by peneplained mountains and hills from the Caledonian and Hercinic chains. The Central Dobrogea appears as a horst compared to

adjacent structural units. The larger part of this unit is made of weakly metamorphosed greenschists which are arranged disconformably over a Mesozoic crystalline layer. The South Dobrogea corresponds to a plateau with the low amplitude, from 60-80 m in the northern part to ~140 m in the south. Large areas are cover by loess, with up to 20 m in the North Dobrogea, to more tens of meters in the South Dobrogea (Posea, 2002).

Figure 3.3. Sketch of southern tip of the Tornquist–Teisseyre Line (modified after Hippolyte, 2002). NDO = North Dobrogean Orogen; CD = Central Dobrogea; SD = South Dobrogea; FB = Carpathian Foreland Basin; HCM = Holy Cross Mts; PCF = Peceneaga–Camena Fault; RRF = Rava– Ruska Fault; SIF = Siret Fault; SOF = Solca Fault.

3.2. Geomorphology

Southeastern Europe is geomorphologically a very diverse region. In the south, it is comprised of the Rhodopes, Balkans and Dinarides (and Pindus mountains if Greece is included). In the north, the Carpathian Mountains separate the Middle Danube Basin in the west, and the Lower Danube Basin in the east.

Orlovat is situated in the southern part of the Middle Danube Basin in the Vojvodina province in Northern Serbia (Figs 3.1). The Vojvodina province is a lowland region, with only two mountains, the Fruška Gora (Crveni Čot 539 m) and the Vršac Mountains (Gudurički vrh 641 m). Moreover, two inland sand fields are present. The Bačka Sands is in the north of Vojvodina and the Deliblato Sands (Deliblatska peščara) is in the southeast of Vojvodina. The Deliblato Sands is an especially interesting feature for the understanding of the past sedimentological conditions at the Orlovat site. It has an elongated and elliptical shape, spreading continuously from the Danube in a north-west direction towards Vladimirovci. The length is approximately 35 km, and the width ranges from 10 to 20 km, and the total area of sand is around 40,000 ha. The directions of the dunes and the depressions between dunes are commonly southeast-northwest, very likely due to the influence of southeast Košava wind (Bukurov, 1982).

Other geomorphological features characteristic for the Vojvodina province are loess plateaus, river terraces and alluvial planes. There are five loess plateaus: Bačka, Titel, Srem, Banat and Tamiš loess plateaus. The Bačka loess plateau surrounds the Bačka Sands (Deliblatska Peščara). The height decreases from northwest to southeast (125 m a.s.l. near Subotica and 86 m a.s.l. near Srbobran). The best-studied sections are Crvenka (Stevens et al., 2011; Zech et al., 2013) and Bačka Topola (Marković et al., 2008).

The Titel loess plateau is an isolated plateau near the confluence of Tisa into the Danube. It covers an area of 80.4 km2 and the highest point is 128 m a.s.l. The Titel loess plateau is well studied in the sense of palaeoclimate (Basarin et al., 2014; Bokhorst et al., 2009, 2011). The Srem loess plateau surrounds the Fruška Gora Mountains. It covers a large area of the Srem region south of Fruška Gora, while the norther limits are fragmented. The Srem loess plateau reaches a height of up to 150 m a.s.l., while the fragments of loess sediment can be deposited at a height of up to 400 m. The best studied sections for palaeoclimate investigations are Ruma (Marković et al., 2006; Vandenberghe et al., 2014) and Irig (Marković et al., 2007) in 25

the interior of the plateau, and Stari Slankamen (Marković et al., 2011), Batajnica (Marković et al., 2009) and Surduk (Antoine et al., 2009) on the eastern bank.

In the Banat region, there are two loess plateaus, Banat and Tamiš. These plateaus are not intensively studied. The Banat loess plateau surrounds the Deliblato Sands, while the Tamis loess plateau is the smallest plateau in Vojvodina. The Orlovat section is located in the Tamiš loess plateau. According to Popov et al. (2008, 2012), the Tamiš loess plateau is a remnant of once a much larger plateau. This is indicated by its location between the floodplain of the Tamiš River (in the northeast, east and southeast) and the palaeochannels Petra and Šozov in the west and northeast. Previous investigations on the Orlovat section were recently conducted (Lukić et al., 2014; Marković et al., 2014a).

River terraces and alluvial fans are the lowest and youngest geomorphological features in the Vojvodina province. They are result of changing dynamics of the big lowlands rivers, such as the Danube, Tisa, Sava, Tamiš, Bega etc.

The Stalać section is located in the Central Balkans. This is mostly a mountainous area, but the section is situated close to the river valley of the Južna Morava River, ca. 3 km from the confluence of the Južna and Zapadna Morava Rivers into the Great Morava (Fig. 3.1). Thus, the surrounding geomorphology is very diverse. However, the transition of the landforms is very gradual; alluvial plains are surrounded by river terraces that gradually transform to a hilly area, and finally to the mountains. Partly, a karst relief is also present. Southwest and especially west parts of this region are higher that the southeast, east and north areas. Loess is preserved only partially in this region, and there are no larger plateaus. Usually, only the last glacial loess is preserved in fragments (e.g. Basarin et al., 2011), and the Stalać section preserves a unique record of several glacial-interglacial phases.

The Urluia and Vlasca sections are located in the Dobrogea region. The Dobrogea is a geographic province of southeastern Romania, which extends NNE–SSW between the Danube and the Black Sea. Palaeogeographical evolution with associated erosion dynamics led to the formation of relief units in the Dobrogea region characterised by low altitude plateau structure. The Dobrogea plateau looks like a relatively rigid plateau, formed on ancient rocks (schists green granite) represented as sedimentary Mesozoic and Cenozoic deposits. Those rocks are heavily eroded by the action of prolonged exogenous factors, forming a softly waved relief with relatively low altitudes (100-300 m). The northern part is 26

higher, reaching in some places at 350-400 m and 467 m in the highest peak (the Greci Macin Mountains). The south is below 200 m (maximum altitude is 204 m at Plateau Deliorman) (Posea, 2002).

3.3. Climate regime

Southeastern Europe lies at the interface between the Continental, Atlantic and Mediterranean climate zones (Fitzsimmons et al., 2012; Stevens et al., 2011). Currently, the Mediterranean influence is limited to Southern Greece and the coast line of the Balkans. It only penetrates deeper into the interior of the Balkans through wider valleys of the larger rivers. The climate of the interior of the Balkans is moderately continental, partly determined by topographic variations where precipitation increases with increased elevation. In the lowland areas the climate is continental, which is manifested with high inter-annual fluctuations in temperature with cold winters and warm, relatively dry summers. However, notable differences can be observed between those regions.

The closest climatic station to Orlovat (Chapter 5) is Zrenjanin (air distance ~20 km), which shows a mean annual air temperature of 11.5˚C, and a mean annual precipitation of 583.2 mm for the period between 1981 and 2010 (Republic Hydrometeorological Service of Serbia). This area is classified as Cbf (temperate oceanic climate) according to the Köppen classification system, although with a strong tendency of Cfa (humid subtropical climate; Sträßler, 1998). A predominant drought period is absent (Fig. 3.4), while the dry period is during August. The warmest months are July and August with an average of 22.2˚C and 21.8˚C, respectively. The winter months are cold, and the mean temperature for January is 0.1˚C. Annual air temperature oscillations are high, and the temperatures for the period between 1981 and 2010 were in the range from 42.9˚C to -27.3˚C. The monthly precipitation maximum is in the late spring (June has an average of 88.8 mm), while the driest month is February (30.0 mm). Average of snow precipitation is 22 days for the period between 1981 and 2010 (Republic Hydrometeorological Service of Serbia).

The Stalać section is situated in Central Serbia and the closest climatic station is Kruševac (air distance ~13 km). For the period between 1981 and 2010 this station recorded a mean annual air temperature of 11.4˚C and a mean annual precipitation of 628.1 mm (Republic Hydrometeorological Service of Serbia). This area is classified as Cbf climate according to the 27

Köppen classification system, without a drought period, albeit with dry summer months (Sträßler, 1998). The summer dry periods are slightly milder when compared to the southern part of the Middle Danube Basin (Fig. 3.4). Similar to the Middle Danube Basin, the warmest months are July (21.8˚C) and August (21.5˚C), while January is the coldest with 0.2˚C. Large oscillations on the annual scale are present, and the temperatures for the period between 1981 and 2010 were in the range from 43.7˚C to -26.0˚C. June is the month with the highest precipitation (71.2 mm), while February is the driest month (39.2 mm) (Fig. 3.4). The average of snow precipitation is higher than in the Middle Danube Basin, and it is 31 days, while the snow cover stays in average 44 days for the period between 1981 and 2010 (Republic Hydrometeorological Service of Serbia). However, it is important to note that in the south of the Middle Danube Basin precipitation is uniformly distributed, while over the Central Balkans, where Stalać is located, precipitation is strongly dependent on the altitude. Therefore, areas of Central Serbia, especially in the west, can have a mean annual precipitation of up to more than 900 mm.

The Urluia and Vlasca sections are in the Dobrogea region in southeastern Romania. The closest climate station is in Constanta (air distance from Urluia ~58 km, and from Vlasca ~68 km), and it recorded a mean annual temperature of 11.5˚C and a mean annual precipitation is 396 mm (for the period from 1961-1990) (World Meteorological Organization, 1996). This region is close to the Black Sea, but due to northerly winds prevailing for most of the year it is a very dry region (Jordanova and Petersen, 1999). This area has Cfa climate, with 6 months of dryness and 3 months of drought (Fig. 3.4.; July, August and September) (Sträßler, 1998). The warmest months are July and August with an average of 22.0 ˚C and 21.8˚C, respectively, and both months are characterised by mean precipitations of less than 33 mm (Fig. 3.4). The coldest month is January with a mean temperature of 0.5˚C.

Figure 3.4. Climate diagrams of Zrenjanin, Kruševac (Serbia), and Constanta (Romania). The diagrams were created on the basis of the climatological normal for the period 1991-2010 for Zrenjanin and Stalać (dotted lines represent a time period from 1961-1990), and from 1961– 1990 for Constanta (World Meteorological Organization, 1996). Periods of drought occur when the mean monthly precipitation curve is below the mean monthly temperatures curve on the diagram (Walter, 1974).

3.4. Hydrology

Large river systems are suggested to be mandatory for loess formation (e.g. Badura et al., 2013; Smalley et al., 2009), and recently it was experimentally demonstrated that the largest loess area in China (the Chinese Loess Plateau) is determined by the Yellow River activities (Nie et al., 2015). The classical research from Smalley and Leach (1978) proposed that the Southeastern Europe loess originates predominantly from the Danube River and it confluences. This was later confirmed by geochemical research (Buggle et al., 2008; Újvári et al., 2008). This is because large lowland rivers, such as the Danube and its tributaries, carry a huge amount of sediment that is occasionally deposited on the river shores. Sediment deposited on that way represent an important sediment supply that can easily be transported by the wind. Therefore, hydrological conditions have played a major role in loess formation. Nevertheless, this also means that the change in the rivers discharge and sediment supply, change in the rivers courses and environmental conditions around the rivers have played an important role for loess formation. Consequently, for a proper understanding of loess records, it is important to understand the hydrological conditions of rivers that have been major source areas. River systems that were probably the most important major source areas for the sections studied in this dissertation were the Danube, Tisa, Tamiš and Južna, Zapadna and Velika Morava Rivers.

The Danube River is the second longest river in the Europe and exceeds 2800 km in length. The present watershed divide between the Danube, Rhone and Rhine Rivers lies in the border region between France, Germany and Switzerland. The stream gradient in the Danube River headwaters, as far as Bratislava on the western edge of the Pannonian Basin is relatively steep. In the vicinity of the Alps, large moraine systems and glaciofluvial terraces formed in response to Quaternary glaciation events (e.g. Marković et al, 2015, 2016). In comparison, the lowland river in the Middle Danube and Wallachian Basins further downstream has a gentler stream gradient. The lowland Middle Danube and Wallachian Basins are filled with thick polycyclic sedimentary deposits of Pliocene and Quaternary age (e.g. Jipa, 2014). Glaciofluvial and fluvial activities in the Danube headwaters have therefore necessarily influenced the timing and nature of sedimentation downstream.

Pekarova and Pekar (2006) provide the long-term trends of yearly discharge time series and runoff variability at seven stations along the River Danube during the period 1901-2000 (with

the only exception for Ceatal Izmail station (Romania), where the provided period is from 1921-2000). The major tributaries (Drava, Tisa, Sava and Tamiš) join the Danube River in the sector between the Drava and Velika Morava confluences to the Danube influenced almost double increase in mean water discharges. At the border between Serbia and Romania which represents 70 % of the drainage area, the Danube already reach almost 90 % of its total discharge.

Besides the Danube, two larger rivers that may have had an influence as the source areas of the Orlovat section are the Tisa and Tamiš Rivers. The Tisa River has a source in the northeast part of the Carpathian Mountains. Its catchment is 157,135 km2 wide. Prior to recovery started far back in the 18th century, the River Tisa was 1429 km long. By multiple cut-offs via 112 larger and smaller meanders its length was shortened by 463 km, and it is only 966 km today (Pavić, 2006). In Vojvodina the Tisa enters from Hungary 6 km downstream from Szeged. The Tisa and Sava Rivers are the longest Danube tributary in the middle part of the river basin. The Tisa water course is regulated by several constructed dams (Svirčev et al., 2009).

The Tamiš River is the second largest river in Banat, with the source in the Carpathian Mountains in Romania and mouth in the Danube near Pančevo or Čenta (during the Danube high tides). The drainage area covers 10,280 km2. With the Danube, it belongs to the Black Sea drainage basin. The river’s total length is 339.7 km (Prohaska, 1998). The majority of its waters come from the Carpathian Mountains. The River Tamiš is in its lower part slow- flowing, chained by embankments and controlled by dams. For the last 53 km it is navigable. The water regime of the Tamiš River has been dramatically changed after the connection with the Danube-Tisa-Danube hydro-system (Marković et al., 1998). The present hydrological characteristics depend on the dam systems in the Tamiš River and the Karašac channel (Prohaska, 1998).

The Stalać section is located at the confluence of the Zapadna and Južna Morava Rivers, where those rivers make the Velika Morava River. Therefore, all three rivers are potential major source areas for Stalać loess. The Zapadna Morava River (West Morava) is a river in central Serbia that originates in the Tašti field, from the Golijska Moravica and Đetinja headstreams. The Đetinja River receives from the left its main tributary, the Skrapež, just less than a kilometer before the confluence. It meets the Golijska Moravica from the south,

forming the Zapadna Morava. Since the proximity of the confluences of the Đetinja, Skrapež and Golijska Moravica, some sources consider all three rivers to be direct headstreams of the Zapadna Morava. Following the direction of the course, the Đetinja is a natural headstream of the Zapadna Morava, but since the Golijska Moravica is 23 km longer, the latter is considered as the main headstream. Measured from the source of the Golijska Moravica, the Zapadna Morava is 308 km long, while the length of the Zapadna Morava without tributary is 210 km. The catchment area is 15,849 km2.The most important confluence of the Zapadna Morava is the Ibar River, which contributes with almost same water discharge at the confluence. Unlike the Južna and Velika Morava's meridian (south to north) flow, the Zapadna Morava runs in a latitudinal (west to east) direction.

The Južna Morava River (the South Morava) is a river in southern Serbia, which represents the shorter headwater of the Velika Morava River. Today, it is 295 km long, and the catchment area is 15,469 km2. It flows generally in the south to north direction. The Južna Morava used to be 318 km long and represented the longer and natural headwater of the Velika Morava. Causing severe floods in history, the meandering river has been shortened by almost 30 km until today, so it became shorter than the Zapadna Morava. However, the Zapadna Morava has always had a higher discharge. Over past few centuries, the Južna Morava catchment has been almost completely deforested, which causes one of the most severe cases of erosion in the Balkans. As a result of this, the river brings a large amount of materials to the Velika Morava, filling and elevating its river bed.

The Velika (Great) Morava begins at the confluence of the Južna Morava and the Zapadna Morava. From there to its confluence with the Danube northeast of the city of Smederevo, the Velika Morava is 185 km long, but together with the Zapadna Morava it is 493 km long. The catchment area of the Velika Morava is 6,126 km², and that of the whole Morava system is 37,444 km². The most distant water source in the Morava watershed is the source of the Ibar River, making the Ibar - Zapadna Morava - Velika Morava river system with a length of 550 km the longest waterway in the Balkans.

3.5. Pedology

The Vojvodina region is a lowland area and the most widespread soil is chernozem. Chernozems are humus-rich grassland soils that are usually found in the middle latitudes of both hemispheres, covering ~1.8 % of the total continental land. Because of high humus content (7-15 %) this soil is very fertile. Chernozems usually form in climatic zones with seasonal rainfall of 450–600 mm per year. They are characterised by steppe and mosaic vegetation; however, the original steppe vegetation is mostly modified into agriculture fields. Investigated loess plateaus in the Vojvodina region are almost completely covered by chernozem soil types.

Pedological characteristics of Central Serbia (the Central Balkans) are more complex due to a stronger gradient in relief and precipitation. The most common are vertisols and cambisols, while fluvisols and podzols are notably present. Vertisols are mostly developed on clay reach substrates of mostly lacustrine origin.

The most abundant soil of the southern Dobrogea is also the chernozem. The distribution of the soil combinations of the central and northern Dobrogea, having chernozems, cambic chernozems, luvisols, rendzines or lithosols as the main component, reflects altitudinal soil differentiation and the sporadic spreading of the massive rocks in the territory.

4. Loess: Rock, sediment or soil - What is missing for its definition?

Tobias Sprafke1, Igor Obreht2

1Institute of Geography and Geology, University of Würzburg, Am Hubland, 97074 Würzburg, Germany

2Department of Geography, RWTH Aachen University, Templergraben 55, 52056 Aachen, Germany

Published 2016 in Quaternary International 399, 208-217

Abstract

Loess is commonly defined as an accumulation of windblown silt. However, the complex mechanisms that are responsible for most of the structural characteristics of loess require a more precise explanation. The common definition of loess ignores a set of processes that start during and after the subaerial deposition of silt. The term loessification has been used by a number of authors to refer to the quasi-pedogenic/quasi-diagenetic processes that result in the typical aggregation of loess; however these mechanisms are rarely described in detail. Depending on the researcher's background, loess is classified as sediment, soil or rock.

This review gives an overview on loess definitions through time and evaluates the main concepts related to loess formation. Several gaps of knowledge are identified that require a number of specific studies related to different aspects of loess formation in various environments. We propose to 1) differentiate primary loessification which initializes the formation of loess structure from secondary loessification which takes place subsequently and 2) define loess-like sediments as deposits that experienced loessification but were not transported aeolian. Mainly aeolian processes and loessification are of equal importance to define loess.

4.1.Introduction

Loess is commonly defined as an accumulation of windblown silt, covering more than 10% of the earth's surface, mainly in the earth's temperate zone (Pye, 1995, Pécsi and Richter, 1996 and Muhs and Bettis, 2003). Loessification is a term that groups the quasi-pedogenic or quasi- diagenetic processes that transfer a loose silt-dominated substrate to a loess-like deposit, with

its specific characteristics (Pécsi and Richter, 1996 and Smalley and Marković, 2014). However, the general understanding of these processes is very limited, hampering a classification of loess as sediment, soil or rock (Smalley et al., 2011). To date, there is still no commonly accepted definition of loess (Muhs and Bettis, 2003), despite its significance.

Since its first description published 190 years ago (Leonhard, 1824), the research on the formation of loess – or substrates that were termed that way – led to numerous attempts to define it genetically (Smalley et al., 2001). The one characteristic found in all definitions is the predominance of silt, especially in its coarse fraction (Pécsi, 1990, Pye, 1995 and Muhs et al., 2014). Smalley (1971) and Smalley et al. (2001) conclude that a major conflict regarding loess definition is related to competing fundamental perspectives of two major disciplines: Sedimentologists stress the processes of material production, transport, and accumulation by wind; pedologists focus on post-depositional processes that form many significant characteristics of loess, but that also take place in substrates that are not completely of aeolian origin (Smalley, 1971).

Any yellowish, carbonate bearing, quartz rich, silt dominated substrate (minor contents of [fine]sand and clay) from an outcrop in Central Europe (e.g. the type locality Haarlass near Heidelberg, Fig. 4.1A), formed by aeolian deposition and aggregated by loessification during glacial times would be widely accepted as (typical) loess. However, many deposits share only a part of these characteristics and defining them as loess may be questionable.

Some authors assume that aeolian deposition of mineral dust is the key process to define loess as a sedimentary body and neglect the significance of other concepts (Pye, 1995, Muhs and Bettis, 2003 and Muhs et al., 2014). However, significant characteristics of many loess deposits are related to post-depositional processes (Pécsi, 1990, Pécsi, 1995 and Smalley and Marković, 2014), such as cementation and aggregation, in geological and pedological terms, respectively. Diagenetic and pedogenetic alterations are mentioned repeatedly to refer to loess as (soft/sedimentary) rock or (synsedimentary/synlithogenic) soil. Frequently, the term loessification is used as a synonym for the post-depositional processes that are relevant to form loess (Pécsi, 1990 and Pécsi, 1995), although it is still not clear which are the geological and ecological factors leading to the specific loess-like structures of many quaternary silt- dominated sediments.

Despite the contemporary trend to reduce loess to a deposit of windblown dust (Muhs et al., 2014), due to recent findings on possible syn-to post-depositional processes (Svirčev et al., 2013) and remaining questions on what loess is (Makeev, 2009), this review focuses on identifying the reasons why it is difficult to come to a commonly accepted definition of loess. First the significance of loess and related deposits is summarised to highlight the reasons behind the different perspectives. Second, an overview of loess definitions through time serves to illustrate how different perspectives developed. The main processes attributed to loess and related deposits are then reviewed and gaps of knowledge for a commonly accepted definition of loess identified.

Figure 4.1. Loess(-like) deposits. A: Haarlass, Heidelberg (Central Europe). Type section of loess (Leonhard, 1824). Photo by C. Hornung (Heidelberg). B: Batajnica, Serbia (SE-Europe). One of few plateau-like loess deposits in Europe with quasi-continuous loess-palaesol sequences. C: Landscape in the surrounding of Luochuan, Chinese loess plateau. Largest loess deposit in the world with longest LPS and specific loess morphology. Photo by Z. Svirčev (Novi Sad), D: Caspian Lowlands, Iran, close to border to Turkmenistan: Small mesas (10-20 cm height) in the semi-desert composed mainly of silty to fine sandy sediments. Mean annual precipitation: ~ 100-200 mm. The darker mesas are connected to cyanobacterial mats which partly prevent the silty parent material from erosion. 36

4.2.Significance of loess and related deposits

Loess and related deposits are one of the most widespread Quaternary sedimentary formations, most abundant in semi-arid regions of inner Eurasia (Smalley et al., 2011). In the humid temperate area, loess deposits formed during the cold stages of the Pleistocene. Alternations of loess and palaeosols are usually interpreted to reflect the broad global climatic oscillations during the Quaternary (Muhs and Bettis, 2003). Though, there are several examples for interglacial loess formation indicated by studies of Holocene deposits (Lehmkuhl et al., 2014 and Muhs et al., 2014), which are often related to (semi-)arid areas (Crouvi et al., 2010). Further confirmations of recent loess come from Alaska (Muhs et al., 2003) and the Chinese Loess Plateau (Pécsi and Richter, 1996). In the Danube basin, where loess-palaeosol sequences (LPS; Fig. 4.1B) rather clearly illustrate environmental changes during the last ten glacial–interglacial cycles (Fitzsimmons et al., 2012), recent research has shown that loess formation may not have ceased at the end of the Pleistocene, but continued during Early Holocene times (Marković et al., 2014a). But loess is not only a Quaternary phenomenon; in China loess formation may have started as early as the Miocene (Guo et al., 2002). There are also late Palaeozoic loess deposits of equatorial Pangaea (Soreghan et al., 2008) and Precambrian loessites in northern Norway (Edwards, 1979) – however altered by significant diagenesis.

Loess sediments, mainly composed of windblown particles, are key palaeoclimatic archives, recording past global atmospheric mineral dust dynamics and palaeoenvironmental changes (Kohfeld and Harrison, 2001, Derbyshire, 2003 and Muhs et al., 2014). LPS illustrate the changes in Quaternary palaeoenvironments on millennial and larger time scales (Rousseau et al., 2002, Bronger, 2003, Antoine et al., 2009, Antoine et al., 2013 and Buggle et al., 2013). Being dominated by silt and usually containing carbonate, loess is the parent material of the world's most fertile soils (Pécsi and Richter, 1996 and Catt, 2001). Numerous Palaeolithic archaeological sites are preserved in loess (Haesaerts et al., 1996, Antoine et al., 2003, Einwögerer et al., 2006, Sinitsyn and Hoffecker, 2006, Boguckyj et al., 2009, Einwögerer et al., 2009 and Iovita et al., 2012) and most European Neolithic settlements are associated to fertile loess areas (Pécsi and Richter, 1996). Loess is a widespread building ground with specific geotechnical properties (Derbyshire and Mellors, 1988, Dijkstra et al., 1994 and Assallay et al., 1997), in which the particle aggregation/cementation plays a major role

(Smalley and Marković, 2014). Depending on these distinct characteristics are also geological hazards in loess areas including mudflows or landslides (Pécsi, 1994 and Derbyshire, 2001).

4.3. Loess definitions through time

Different definitions of loess were given through time (Smalley et al., 2001), starting in the early 19th century, when it was conceived as a purely geological object (Leonhard, 1824 and Lyell, 1834). At the turn of the century, scientists from the emerging field of pedology became interested in the complex nature of loess (Berg, 1916). Quaternary geologists used it as a stratigraphic unit representing Quaternary glacials (Ložek, 1965). Continuously, loess had to remain a simply to define mappable geologic body (e.g. windblown silt) and in the late 20th century it was identified as key record of past dust dynamics (Pye, 1995 and Muhs et al., 2014). It is nowadays increasingly recognised as a product of rather complex environmental processes that include both, sedimentary and post-depositional processes (Svirčev et al., 2013). Essentially, loess is a polygenetic entity sharing properties of sediments, soils and rocks. Some definitions are either lengthy (Pécsi, 1990, Pécsi, 1995 and Koch and Neumeister, 2005), others are very short (Pye, 1995 and Muhs et al., 2014), but a broad consensus still seems to be lacking.

4.3.1. Leonhard (1824) and Lyell (1834): a soil-like sediment of alluvial origin

The outcrop Haarlass near Heidelberg (Fig. 4.1A) is recognised as type locality for loess since its first scientific description in “Charakteristik der Felsarten” (Character of rock types) by Leonhard (1824). Loeß or Loesch (Alemannic for loose) are given as synonyms of Schneckenhäusel-Boden (Mollusc-Soil) or Mergel (Marl). Leonhard describes the loamy character of the material, breaking down to soil-like aggregates and easy to pulverize to a dusty mass. From this perspective, next to the predominance of silt, the particular aggregation and the carbonate content became important criteria. Since the definition of Leonhard (1824) and the popularization of the term loess by Lyell (1834), who assumed an alluvial origin for the silty material, research on the genesis of loess and comparable substrates (silt dominated or with loess structure) were carried out and definitions focused more on the formation processes.

4.3.2. Richthofen (1882): aeolian origin of silt particles

According to Pécsi and Richter (1996) it was Virlet D'Aoust (1858) to first publish the idea of loess being an aeolian sediment. Having in mind the widely accepted notion of a fluvial origin of loess, Richthofen made detailed observations on the famous loess plateau of Northern China (Fig. 4.1C) and accepted the aeolian origin for the silt (Richthofen, 1882). This is supposed to be the classical aeolian theory on loess formation, but the descriptive list how loess is characterised does not include information on the processes responsible for its formation (Smalley et al., 2001). Some authors point out that F. v. Richthofen's view was much more than purely sedimentological, acknowledging post-depositional processes and the importance of the environment of deposition (Russell, 1944 and Pécsi and Richter, 1996).

4.3.3. Berg, 1916 and Berg, 1964): in situ formed soil (regardless of parent material)

The substantial finding of Dokuchaev (1883), that one parent material in different climates alters to different soils during the same time, can be seen as the background for Berg (1916) to postulate that loess is a product of weathering/pedogenesis (regardless of the parent geology). In this perspective, loess and loess-like rocks result from the same process. The term loessification is frequently attributed to L. S. Berg or R. J. Russel (Smalley, 1971, Smalley et al., 2010 and Smalley and Marković, 2014); it refers to the development of a loess structure through different processes that take place in the ground. These are, however, not explained in detail. Russell (1944) brought the in situ theories to loess-like sediments in the USA (Mississippi Valley), the heartland of the sedimentological perspective on loess (Smalley, 1971).

4.3.4. Ložek (1965): complex formation of loess environment

Ložek (1965) attempted to balance both the aeolian and in situ processes to explain loess formation. According to the author, both aspects are equally important and reflect the “loess environment”. Following Ložek (1965), the molluscs in loess deposits constitute a discrete biostratigraphic unit that is different from all other quaternary mollusc assemblages. The mollusc assemblage of loess deposits has no present day analogue and represents cold steppe to tundra environments. From this standpoint, Ložek (1965) stresses that, since the term loess derived from a Central European deposit, it can only refer to glacial phases in this region. Thus, loess in the semi-arid regions of Asia may not be covered by this definition.

Until the late 1960s, loess deposits were a major reference for Quaternary stratigraphy (Fink, 1961, Fink, 1964 and Semmel, 1968). LPS were seen to reflect landscape response to climate changes during glacial–interglacial cycles. One major loess researcher and close colleague of V. Ložek was G. Kukla, who first correlated LPS with marine sediments (Fink and Kukla, 1977 and Kukla, 1977). He later worked on the Chinese Loess Plateau, the thickest and largest loess deposit in the world (Kukla, 1987, Kukla et al., 1988 and Kukla and An, 1989). Kukla (1987) follows the idea that loess is not only a product of aeolian sedimentation, but also requires specific pedogenic processes (e.g. loose cementation by carbonate incrustations); however, he relates these processes mainly to a semi-arid continental climate.

4.3.5. Liu (1988): more than a deposit of dust storms

Richthofen (1882) realised striking lithological similarities between loess deposits in Central Europe (Fig. 4.1B) and the extraordinary thick loesses of China (Fig. 4.1C). “Huangtu” (yellow earth) is the Chinese name for these loose earthy deposits. Reports, as old as 2000 years, on significant dust storms indicate that an aeolian origin of Chinese loess was understood quite early. Unfortunately for Chinese scholars, the merit to first publish these ideas in western journals is attributed to F. v. Richthofen (Liu, 1988). Accepting the evidence for aeolian processes as significant elements of loess genesis in China, Liu (1988) also points to a final biochemical stage of loess formation, which refers to the post-depositional processes taking place after the dust settles. In arid climates with low biological productivity, secondary carbonates form rapidly, whereas in more humid environments other weathering processes are dominant and lead to pedogenesis. Secondary carbonates are mentioned to play a significant role to make loess being a buff, porous and homogenous earth – the process is termed loessification. However, Liu (1988) points out that he disagrees with the theories of Berg (1964) who totally neglects the importance of aeolian processes for loess genesis.

4.3.6. Pye (1987 and 1995): windblown silt deposit

In contrast to the theory of Berg (1916), inspired by classic Russian pedology, most American and British scientists throughout the second half of the 20th century based their work on the sedimentological perspective of loess genesis (Muhs et al., 2014). Pye stated that loess is just an accumulation of dust (Pye, 1987, Pye, 1995 and Pye and Sherwin, 1999), a definition that found wide support for explaining the processes behind the accumulation of large amounts of silt sized quartz (Smalley, 1971). The production of silt-sized quartz by glacial grinding was 40

long seen as the crucial process in loess genesis, although in recent times other processes with less intensity but large geographical distribution are accepted (Soreghan et al., 2008). There are many attempts to clarify the pathway from primary silt to the final deposit of loess or loess-like sediments, including a complex chain of transport agencies, like rivers and wind (Wright, 2001a and Smalley et al., 2009). Thus, having explained most of the sedimentological aspects of loess formation, it was widely assumed that these may be sufficient to give a simple definition of loess (Smalley et al., 2011 and Muhs et al., 2014). However, it may have not been intended that loess and dust are used as synonyms and silt components in soils and sediments are termed loess. Iriondo and Kröhling (2007) see loess as fine-grained sediment deposit of rather universal occurrence. The authors advocate the presence of tropical loess and other varieties to challenge the classic view of loess being linked to (peri)glacial environments. Desert loess, dust produced and transported from deserts is a concept that experienced controversial debate since decades (Smalley and Vita-Finzi, 1968, Yaalon and Dan, 1974, Whalley et al., 1982, Wright, 2001a, Wright, 2001b and Crouvi et al., 2010). The vision of loess as mainly aeolian dust made loess deposits attractive as archives of past global and regional mineral dust dynamics, a topic that found increasing interest in climatology during the last decades (Kohfeld and Harrison, 2001, Rousseau et al., 2002 and Muhs et al., 2014).

4.3.7. Pécsi and Richter, 1996 and Pécsi, 1995): not just the accumulation of dust

The famous publication of Pécsi (1990) entitled “Loess is not just the accumulation of dust” and further works by the author (Pécsi, 1995 and Pécsi and Richter, 1996) can be regarded as the last considerable pleas to an in situ theory of loess formation and definition. In previous years, several publications (e.g. Pye, 1987) contributed to the general acceptance of the sedimentological perspective. As a reaction, Pécsi (1990) described in detail the characteristics of typical loess, and suggested at least ten descriptive criteria for this substrate, possibly referring to Richthofen's list, but including post-depositional processes that take place in certain environments. With reference to Russian authors, Pécsi (1990) postulates the superzone of loess formation as desert margins, (forest-)steppe, and (forest-)tundra. Loessification as a central process of loess genesis is only mentioned in the abstract of this publication, but not in the text; neither is given an explanation of the processes. Pécsi (1995) identifies Fe and Mn films, dispersed limonite, carbonates and humus as cementing agents.

Controversial opinions about how loess is to be defined, identified in the field and finally mapped find their visual remainder in the European loess map of Haase et al. (2007). Almost exclusively in Europe, large parts of Hungary, homeland of M. Pécsi, are covered by “alluvial loess” (Fig. 4.2).

Figure 4.2. Central part of the European loess map of Haase et al. (2007) with abbreviated capital names. Localities mentioned in Figs. 4.1 and 4.3 and political boundaries of Hungary indicated. Note the rather exclusive appearance of alluvial loess in this country's area.

4.3.8. Recent developments

It is nowadays accepted that aeolian processes are a main element in loess genesis (Muhs et al., 2014). However, in situ processes are gaining more recognition and loess deposits are being increasingly studied as complex palaeoenvironmental archives. The environment of dust deposition and the process of loessification are the subject of recent studies (Smalley et al., 2011 and Svirčev et al., 2013). Biological loess crusts (BLC) are investigated as one possibly crucial factor in loess formation in arid and semi-arid areas, where they develop on silty substrate and collect and preserve dust particles (Svirčev et al., 2013; cf. Fig. 4.1D). Smalley and Marković (2014) also stress the relevance of post-depositional processes in providing loess with its technical properties.

Koch and Neumeister (2005) suggest a broad terminology (in German) to classify loess according to its genesis. In this classification, typical loess is an aeolian deposit that is loessified – a term not explained in detail; and loess derivates are aeolian deposits modified during or after deposition. Besides these two, the authors also suggest a number of subtypes to loess and loess derivates. Sprafke et al. (2013) use the term solifluction loess (Koch and Neumeister, 2005) to describe the loess sediments of Paudorf locus typicus (Lower Austria), a yellowish, wand-stable substrate rather rich in carbonate (10–25 %) with loess-like structure, but containing gravel throughout the sequence (comparable deposits can be found in the whole region around Krems; Fig. 4.3). The authors assume a seasonal deposition of silt, accompanied by solifluction, admixture with rock fragments and loessification in cold steppe to tundra environment as main agents in the formation of these deposits. However, as the final transport was colluvial and not aeolian, the deposit may not be named loess, despite its structure and overall loess-like appearance. It is possibly comparable to “derasion loess” (Pécsi and Richter, 1996), which may be rather widespread throughout many loess regions. Most loess deposits outside plateaus depend on distinct morphological positions, which are often slopes on the leeward side to predominant wind directions.

Figure 4.3. Polygenetic loess in the Krems region, Lower Austria, possibly comparable to derasion loess of Pécsi and Richter (1996). A: Morphological situation of Krems Schießstätte (white rectangle). Location at E-exposed slope (i.e. leeward postion to W-winds) on a spur between Danube and Krems River. B: View of the outcrop Schießstätte: Mostly a wall-stable yellowish substrate dominated by silt, i.e. a loess-like sediment. C: Polygenetic loess from the northern part of the outcrop. Note the distinct gravel bands (partly discontinuous) and the presence of gravel throughout the sequences indicated by small holes due to profile cleaning (some marked). D: Aggregates of loess-like sediments from this profile. Upper one weakly structured, lower one with slight biogenic structure (size of aggregates: 4 x 2.5 cm).

4.4.Loess genesis and the role of loessification

4.4.1. What is loessification?

Loessification is considered a significant process in loess genesis (Berg, 1916, Berg, 1964, Russell, 1944, Ložek, 1965, Liu, 1988, Pécsi, 1990, Pécsi, 1995, Pécsi and Richter, 1996, Cilek, 2001, Koch and Neumeister, 2005, Smalley et al., 2011, Sprafke et al., 2013, Svirčev et al., 2013, Smalley and Marković, 2014 and Sprafke et al., 2014). Despite different views on what loessification is, several authors agree that loess is “not just the accumulation of dust” (Pécsi, 1990). Loessification distinguishes loess from mineral dust deposits and is crucial for the development of loess(-like) sediments. Loessification provides sediments, especially silt dominated, with the typical loess structure and appearance. While authors agree on the result of loessification, theories or studies to explain this process are scarce (Cilek, 2001). Depending on the perspective of loess definition (soil or rock), loessification processes are defined within pedogenesis or diagenesis. Pécsi (1990) uses the terms slight and moderate diagenesis as well as pedological next to physical, chemical and biological processes when pointing to loessification (Pécsi, 1995).

The process of loessification would start when dust particles are trapped and protected from further transport, e.g. in steppe or tundra vegetation environments. BLC may play a key role in this process (Svirčev et al., 2013). Loessification is not classified as a pedogenic process (its impact is too weak to actually form a soil), and its relation to the pedosphere is still a matter of debate. However, loessification processes take place close to the ground surface in certain ecosystems, so they are also not purely lithospheric. The relation of loessification and diagenesis is controversial, so we may only be able to speak of quasi-pedogenic and/or quasi- diagenetic processes. This suggests that there are smooth transitions between different spheres where loessification and related processes take place. Further research is necessary to define the boundaries.

Pécsi and Richter (1996) state that loess formation takes place in (semi-)arid environments, (forest-)steppe or tundra ecosystems. According to Pécsi (1990), loessification is a universal process, happening in all these ecosystem types. However, (forest-)steppe ecosystems produce significant amounts of biomass, which leads to rather quick pedogenesis (on the studied time scale). In fact, it is still not well understood, if in different environments the same type of loessification takes place. 45

It is necessary to distinguish initial (primary) loessification from secondary loessification. Initial loessification is the process that initializes the formation of loess-like structure by cementing and/or stabilizing dust particles; i.e. the moment when mineral dust is made to loess after its deposition in a certain environment. The loess-like structure continues to develop in the pedosphere/lithosphere during changing environments, which we refer to as secondary loessification. This takes place as long as the substrate is not altered significantly, e.g. by pedogenesis, redeposition or strong compaction/diagenesis. If environmental conditions remain constant, primary loessification takes place at the ground surface and secondary loessification below. However, the detection of primary/initial loessification may not be straightforward, since secondary loessification may mask the primary stage. We assume that primary loessification may diverge in different ecosystems of loess formation.

To achieve a clear picture of primary and secondary loessification, it is necessarily to evaluate the complete sequence of deposition and alteration processes. For this, further studies in different environments of loess formation are still necessary.

4.4.2. The relevance of carbonate and clay for loess formation

Russell (1944) and Pécsi (1990) claim that the carbonate content is one of the most essential properties of loess, and a significant aspect for the in-situ theories of loess formation (Smalley et al., 2001). According to this view, carbonate-free loess should not be considered as loess but as loamy sediment with similar properties to loess. Smalley (1971) mentions loess and loess-like deposits in New Zealand and Nebraska that are very poor in carbonate, and therefore would not fit to a restricted definition demanding significant amounts of these minerals. However, both deposits were formed during glacial periods and have a loess-like structure. It remains to be understood if the deposits are poor in carbonate content due to their leaching or an absence of these minerals in the parent material. Depending on how these deposits developed, we understand if carbonate is a crucial part of loess genesis and if so, which quantity is indeed significant for this classification.

Besides the parent material, other sources of carbonates in loess could be: hydrocarbonate- bearing groundwater; weathering of calcareous shells; and the activity of microorganisms, even if the parent material contains no detrital carbonate (Pécsi, 1990). Cilek (2001) notes that clay bridges are common in loess and often appear impregnated with calcium carbonate. Secondary calcite may be quickly released from carbonate bearing ground water by freeze– 46

thaw cycles (Cilek, 2001). However, this is not easily extrapolated to drier environments of loess formation, where weathering of calcareous shells may be too slow to cause initial loessification. The activity of microorganisms appears as the only remaining option, however, their influence is yet poorly understood. To date, carbonate from parent material seems to be the only major source of calcite cement.

If the provenance area of a loess deposit does not include carbonate rocks, can far-travelled dust contain sufficient carbonate to cause loessification or is it possible to have loess formation from carbonate-free deposits? Carbonates are assumed to be fundamental for cementing typical loess, but is this taking place during primary loessification? Short-scale translocated clay minerals that are mainly connected to dust particles may be capable of stabilizing the silt particles as well (Cilek, 2001), but such process has not been described yet. It is possible that some environmental conditions favour the formation of a loess-like structure just by stabilizing clays when carbonates are absent. For instance, both the New Zealand and Nebraska loess are reported to be enriched in clay minerals (Smalley, 1971).

4.4.3. Sedimentary processes: the prerequisite for loessification?

There is much less controversy around the sedimentary processes involved in loess formation than the in situ alteration. In this section we summarize the main sedimentary agents described for loess formation, but not the complex sedimentary process involved in the formation of loess-like deposits.

Smalley, 1990 and Wright, 2001b and Soreghan et al. (2008) summarize processes responsible to produce silt sized quartz. The theories of J. Hardcastle and P. A. Tutkovskii in the end of the 19th century are the first to state the importance of glacial grinding to produce quartz dust (Smalley et al., 2001), a view supported by various authors (Smalley, 1966, Boulton, 1978, Smalley and Krinsley, 1978 and Smalley, 1990). However, Smalley (1995) and Assallay et al. (1998) support that physical weathering of the periglacial area or (semi-) deserts contributes to mineral dust production (Goudie et al., 1979, Wright, 2001a and Wright, 2001b). These processes may be weaker than glacial grinding, but the areas where they take place are larger (e.g. periglacial areas, semi-deserts, deserts). Deserts are not only storages of mineral dust (Smalley and Krinsley, 1978), but are also potential producers. For example, the “warm loesses” of Obruchev (1945), which may have formed before the onset of

vast glaciations. However, glaciers and glacial times are still crucial for the production of large dust deposits (Soreghan et al., 2008).

Silt is much easier to be mobilised than clay and sand (clay has higher cohesion and sand higher weight). Dryness and gaps in the vegetation cover, both climatically and morphologically induced (Zech et al., 2013), favour the mobilization of the silt fraction by wind. For Central Europe, it is long accepted that periglacial fluvial deposits, and especially fluvioglacial terraces, were the source of the silt that was mobilised by aeolian action (Weidenbach, 1952). Besides this, dust was also entrained from (semi-)desert areas. For instance, Sahara dust contributes to the finer silt fractions in LPS of the temperate belt (Stuut et al., 2009).

To transport silt to the final place of deposition complex sequences of fluvial and aeolian processes have to be taken into account (Wright, 2001a and Smalley et al., 2009). Both transport agents may produce silt-sized mineral particles. Loess has regional differences in mineralogy, but there is always a global dust component in the finer fraction. Especially during interglacials, e.g. the Holocene, the sources of mineral dust are mainly (semi-)desert areas. In this respect, studies on the grain size distributions of loess deposits are done to estimate the distance travelled by the dust from its source area, and also to indicate palaeowind strength (Pye, 1995, Antoine et al., 2001 and Vandenberghe, 2013).

In loess theories, the final deposition of the silt-sized particles is aeolian by definition. Following Pye (1995), the silt accumulation takes place in locations of reduced wind speed (e.g. leeward sides of slopes) or with physical obstacles (e.g. vegetation). The different processes responsible for trapping the particles into the sediment belong to the transition with the in situ theories of loess formation.

4.4.4. Particle trapping: the onset of loessification?

The factors responsible for trapping the silt or dust particles in loess formation are still to be determined. In this respect, it is important to understand whether loessification is restricted to specific grain sizes, since no loessified clays or sands are reported. Although it is accepted that grain size distribution is influenced by wind strength (Antoine et al., 2009, Bokhorst et al., 2011 and Vandenberghe, 2013), it is never considered that the particle-trapping pattern may favour distinct grain size fractions.

Three main mechanisms are responsible for trapping dust particles (Svirčev et al., 2013): deposition with water droplets, a process that seems to not have been significant during the commonly dry glacial conditions; dry deposition, which refers to accumulation after the reduction in the energy of the transport agent; trapping due to local boundary layer conditions. Specific types of vegetation are regarded as key agents for trapping dust (Tsoar and Pye, 1987). However, it should be noticed that the biomass production is significantly reduced during glacial periods, when the growing season is shorter. A lack of vegetation is not only problematic for the accumulation, but also for the preservation of the deposit. Exposition of the bare surface will result in significant wind erosion (Sweeney and Mason, 2013). Svirčev et al. (2013) propose that BLC can resist averse and dry conditions, and even with a minimum of precipitation they can trap the dust particles and preserve the surface, even during dry periods. Zaady and Offer (2010) observed this pattern in the Negev desert nowadays, where biological soil crusts, mostly made from cyanobacteria, collected wind-blown particles and protected soil from erosion even under conditions with annual precipitation of only 87 mm (cf. Fig. 4.1D) – a pattern comparable to the BLC model. For the nival zone of the high-mountainous areas of the Tian Shan, BLC form on seasonal snow cover and are not only retaining but also aggregating dust particles and increase their resistance to erosion (Glazovskaya, 1952 and Shatravin, 2007). The influence of BLC on primary loessification during glacial conditions is still not well understood and requires more studies, but a combination of vegetation with microbiological cover seems to be the best explanation for the trapping and preservation of the particles.

4.4.5. The problem of polygenetic loesses

A problematic issue related to loessification are the loess-like deposits, where the last agent of particle transportation was not aeolian, e.g. due to slope processes in morphologically active positions (Fig. 4.3A). The derasion loess of Hungary (Pécsi and Richter, 1996) or the solifluction loess in the Krems region (Sprafke et al., 2013) are some examples of loess-like deposits, which may question straightforward models of particle trapping, primary and secondary loessification. The polygenetic loesses of Krems are exposed as carbonate bearing wand-stable substrates and have a loess-like structure, but contain a constant admixture of gravel and show distinct gravel bands (Fig. 4.3B–D). In these cases, a more complex sequence of processes accounts for their formation, where seasonality or other short scale climatic oscillations must be taken into account. Crucial questions revolve around the 49

sequence of dust trapping, the redeposition of particles, and the final development of a loess- like structure. If dust is only trapped in certain ecological conditions, when does redeposition take place? And why is there still a loess-like structure in the final deposit? (Fig. 4.3D). Continuous gravel admixtures (e.g. the LPS Paudorf) could be explained by very strong winds, together with aeolian silt deposition, and synchronous primary loessification. Distinct gravel layers may have formed due to slope processes, but could have been overprinted by secondary loessification processes. Further research is needed to unravel the genesis of polygenetic loesses.

4.4.6. Diagenesis

There are significant differences in the aggregate stability of last glacial, Middle Pleistocene and Early Pleistocene loess deposits (e.g. the loess of the Krems region, still not studied in detail). It remains to be evaluated the degree in which aggregate stability is related to the physical compaction induced by the overlying deposits and to physiochemical processes, such as remobilization/recrystallization of carbonates. A key issue in this respect is where to draw the line between loessification and diagenesis. Can these processes simply be defined as occurring during loess formation of the following glacial period? In the case of Permian siltstones of Central to Western USA, which are interpreted as Pangaean loess-palaeosol sequences (Soreghan et al., 2008), the term diagenesis may be better suited than in any Quaternary loess deposit.

4.4.7. Pedogenesis

Despite the multiple existing ways to define a soil (Semmel, 1993), no attempt was made to distinguish between loessification and pedogenesis. Both process groups may be related to the interaction of sediments with the ecosystem but it is unclear if loessification are only weak pedogenic or different processes. For Precambrian lithological units, the involvement of microorganisms in their formation is regarded as sufficient to name a unit palaeosol (Sheldon and Tabor, 2009). For Quaternary, however, the presence of higher plants should be seen as a decisive element of pedogenesis. The relation between higher plants and microorganisms (e.g. cyanobacteria) with loessification remains unknown. Loess often contains plant root channels and a biogenic structure (Fig. 4.3D), meaning that the formation of loess can also be regarded as belonging to the pedosphere. It is to be asked whether plants produce too much biomass relative to dust sedimentation, which would result in pedogenesis and not loessification. Do 50

the macropores in loess, commonly attributed to higher vegetation (e.g. grasses), form syngenetically? Or do they reflect an environment slightly different to the environment of primary loessification? High sedimentation rates can indeed mask past evidences of pedogenesis (Stevens et al., 2011). However, we assume that even high sedimentation rates cannot completely mask soil formation during periods of high biological production (e.g. interglacials; Obreht et al., submitted for publication). Furthermore soil formation may not be significant in low accumulation environments if there is not sufficient biomass. More information on the length of the growing season and biomass degradation during the Pleistocene is needed to better understand the relationship between loessification and pedogenesis.

4.5.Dust, loess, and loess-like sediments

Most contemporary definitions agree that loess is an aeolian accumulation of mainly silt-sized particles (Muhs et al., 2014). Pécsi (1990) notes that silt-sized particles (10–50 μm) make up on average 50–70 % of loess, but that deviations from this numbers may occur. Hence, definitions based on grain size seem insufficient. Chemical and mineralogical definitions (e.g. based on the presence of a certain component or on the carbonate content) are often limited, because they may include substrates that are far from the specific characteristics of typical loess. Since the widely accepted definition of loess emphasizes on the physical process of its genesis, we propose that a more embracing definition should include both the process of aeolian accumulation and the process of loessification. The definition could be more elaborate than replacing very few words (e.g. accumulation of dust, wind-blown silt deposit) by one word (i.e. loess). However, loessification remains poorly understood, although the concept includes all the processes that can turn the primary particles into a typical loess-like structure. Although loess-formation may vary among different environments, consensus is that mainly silt-sized particles are prone to loessification and that the environments of loessification are often characterised by aeolian processes.

Finally, loessification defines both loess and loess-like sediments, and aeolian origin distinguishes loess from loess-like sediments. As the aeolian genesis of loess is considered as crucial, it remains still difficult to distinguish whether a substrate is purely or only mainly aeolian in origin. Polygenetic loesses, from Hungary or the Krems region in Austria, may only be few examples of loess-deposits that are related to morphologically active slope

positions. Therefore, a definition based only on pure aeolian accumulation seems too exclusive. On the other hand, if loess can be assumed as mainly aeolian accumulated sediment, where do we draw the line between loess and loess-like sediments? From a genetic standpoint, it may be necessary to define a complex genesis that includes other sedimentary processes, in addition to aeolian deposition, and talk about polygenetic loess or loess-like sediments more often. However, if the sedimentary structure allows the reconstruction of palaeowind-dynamics and dust flux during loess formation we may consider it as loess.

Loess may be defined as mainly aeolian accumulated dust that has been exposed to loessification (having a loess-like structure). Loess-like sediments, in contrast have a loess- like structure but are polygenetic by definition. They may contain considerable depositional hiatuses. They can provide with information on past environmental conditions, but cannot be used for detailed dust flux reconstructions. Loess derivates are loess or loess-like sediments that have been altered and do not show a loess-like structure. The umbrella term for loess, loess-like sediments and loess derivates could be loess sediments after Koch and Neumeister (2005).

The final question on whether loess is sediment, soil or rock, demands a proper understanding of what the process of loessification is. If we discover that loessification is different from diagenesis, then loess is not a sedimentary rock. If loessification turns out to be different from pedogenesis, then loess is not a (synsedimentary/weak) soil. However, if loessification implies both diagenesis and pedogenesis, then loess cannot be a pure sediment (dust). Loessification is the crucial element of loess genesis and provides the deposits with essential characteristics. Until we gain a better understanding of what loessification actually is, then the best way to classify loess is not by naming it a rock, a soil or a sediment, but simply assume it as loess.

4.6.Conclusions

The significance of loess and related deposits is widely accepted, although there is no commonly accepted definition of loess, and no consensus if it is rock, soil or sediment. The review of loess definitions though time shows various foci on loess definitions that partly reflect broader scientific trends. From the sedimentological perspective many questions regarding loess genesis appeared to be rather solved in the last decades. However, an overview of the processes relevant for loess genesis has revealed many open fields for future research, which are mainly related to the concept of loessification. The role of particle trapping in different ecosystems, the further processes included in maintaining, strengthening or destroying the loess-like structure, and their differentiation from pedogenesis and diagenesis require rigorous studies. To date, we can only state that loess is neither sediment, soil, nor rock, but something complex in between, and that loessification and related processes need to be understood to give a satisfying loess definition. For now, we propose that loess is a mainly aeolian dust deposit that experienced loessification, resulting in loess- like aggregation. Loess-like sediments have a more complex genesis, but are structurally similar to loess. Loess derivates may form of loess or loess like sediments and lack loess-like structure. Loess sediments, as proposed by Koch and Neumeister (2005) can be used as umbrella term.

5. Aeolian dynamics at the Orlovat loess–palaeosol sequence, northern Serbia, based on detailed textural and geochemical evidence

Igor Obreht a,*, Christian Zeeden a, Philipp Schulte a, Ulrich Hambach b, Eileen Eckmeier a, Alida Timar-Gabor c,d, Frank Lehmkuhla a Department of Geography, RWTH Aachen University, Templergraben 55, D-52056, Aachen, Germany b Chair of Geomorphology, Laboratory for Palaeo- and Enviro-Magnetism, University of Bayreuth, D-94450 Bayreuth, Germany c Faculty of Environmental Science, Babeş-Bolyai University, Fântânele 30, 400294 Cluj- Napoca, Romania, and Interdisciplinary Research Institute on Bio-Nano-Science of Babeş- Bolyai University, Treboniu Laurean 42 400271, Cluj-Napoca, Romania d Interdisciplinary Research Institute on Bio-Nano-Science of Babes-Bolyai University, Treboniu Laurean 42, 400271 Cluj-Napoca, Romania

Published 2016 in Aeolian Research 18, 69-81

Abstract

Previous investigations showed that the Orlovat loess–palaeosol section, northern Serbia, is characterised by irregularities in sedimentological properties, magnetic susceptibility and color of the sediment. Here, we applied granulometric analysis and X-ray fluorescence (XRF) analyses to study how the sedimentation at the Orlovat site was conditioned by specific geomorphological or climatic conditions. Grain-size analysis is an established method and one of the most frequently used palaeoenvironmental proxies of loess deposits, and is complemented here with high resolution XRF analysis on sand-free samples to obtain a more detailed insight into palaeoenvironmental conditions and weathering during the past 160 ka. The geomorphological conditions of the surrounding area and variations in wind speed over time are of great importance for a better understanding of loess–palaeosol deposits. The Orlovat section was exposed to special depositional conditions, which differ from other sections studied in the Carpathian Basin. Sand was delivered during interglacials, most probably from the Deliblato Sands by the southeast Košava wind. This study highlights the

importance of an integrated sedimentological approach for reliable palaeoenvironmental reconstruction.

5.1. Introduction

Loess–palaeosol sequences (LPS) have been recognised as sensitive records of past climatic and environmental change (e.g. Derbyshire et al., 1997; Kukla, 1977; Smalley et al., 2011; Stevens et al., 2008). Distribution and origin of loess can give important information about the palaeowind direction (e.g. Muhs and Benedict, 2006; Pye, 1995), and also palaeowind strength can be reconstructed by grain-size distributions (e.g. Vandenberghe and Pissart, 1993; Vandenberghe et al., 1998; Xiao et al., 1995; Yang and Ding, 2014). LPS are commonly observed in the vicinity of the Danube River and present one of the most valuable terrestrial archives in Southeastern Europe (e.g. Buggle et al., 2009). Smalley and Leach (1978) reviewed the origin and distribution of Danubian loess and suggested that loess in the Middle and Lower Danube region originates predominantly from alluvial deposits of lowland rivers, specifically the Danube itself and its tributaries. Recently, the geochemical investigations of Buggle et al. (2008), Újvári et al. (2008) came to the same conclusion.

LPS in the Vojvodina, northern Serbia, have come into focus of European Pleistocene palaeoclimate and loess research during the last years. Previous studies in Serbia used stratigraphy (Basarin et al., 2014; Marković et al., 2008, 2012a), magnetic properties (e.g. Basarin et al., 2011, 2014; Buggle et al., 2009; Liu et al., 2013; Marković et al., 2009), grain- size variations (Antoine et al., 2009; Bokhorst et al., 2009, 2011; Marković et al., 2006, 2007; Vandenberghe et al., 2014; Zech et al., 2013), iron mineralogical (Buggle et al., 2014) and malacological proxies (Marković et al., 2004, 2014b) as well as AAR (Amino acid racemization) relative (Marković et al., 2005, 2011) and absolute luminescence geochronology (Fuchs et al., 2008; Schmidt et al., 2010; Stevens et al., 2011; Timar-Gabor et al., 2015) to investigate palaeoenvironments. All mentioned studies focused on the central, northern and western part of the Vojvodina in Serbia. The eastern part of the Vojvodina, known as the (Serbian) Banat region, has been hardly investigated for palaeoclimatic and palaeoenvironmental changes. However, some archeological findings in the Serbian and Romanian part of the Banat region (e.g. Băltean, 2011; Hahn, 1977; Kels et al., 2014; Sitlivy et al., 2012) indicate the presence of modern humans (Homo sapiens sapiens) during the Upper Palaeolithic, imbedded in terrestrial sediment. For this reason, special attention is paid

to the Orlovat section, the only investigated section in the Serbian part of the Banat region (Lukić et al., 2014; Marković et al., 2014a).

The Orlovat section is situated on the edge of the Tamiš loess plateau (Fig. 1). Popov et al. (2008, 2012) proposed that the Tamiš plateau is a remnant of a once much larger plateau. Besides its palaeoclimatic value, the Orlovat section is also interesting from a sedimentological point of view due to its situation at the slope of the plateau. In previous studies based on rock magnetic (Marković et al., 2014b) and color (Lukić et al., 2014) analysis, it was observed that this section has a unique sedimentology differing from all other investigated sections in Serbia. However, these studies could not answer whether these differences are due to different geomorphological conditions, slope sedimentation processes (e.g. redeposition of older reworked soil material) or systematic differences in climate evolution between the Banat region and the central, western and northern Vojvodina.

The aim of this study is to investigate if sedimentation and weathering processes at the Orlovat section were determined by specific geomorphological, environmental or/and climatic conditions. Special attention is attributed to the surrounding geology and geomorphology. Grain-size analysis and X-ray fluorescence (XRF) analysis are applied in this study, aiming at a better understanding of sedimentological processes. Textural analysis is a well-established method and one of the most frequently used palaeoenvironmental proxies of loess deposits (Ding et al., 2002; Prins et al., 2007; Vandenberghe et al., 1997; Yang and Ding, 2014). In contrast, geochemical analyses constitute a relatively novel approach in the Vojvodina region (Bokhorst et al., 2009; Buggle et al., 2008, 2011, 2013; Zech et al., 2013) and previous studies applied only a low resolution of geochemical proxy values. Since this study generates high resolution geochemical element data for the first time in the region, one of its aims is to give better insight into the mutual dependence of elements within this section. Elemental data is also used to obtain a more detailed understanding of weathering during the past ~160 ka, and to evaluate the reliability of weathering indices.

5.2. The Orlovat site in northeastern Serbia

5.2.1. Study site and geomorphological setting

The investigated LPS is exposed at a brickyard in the village of Orlovat (45˚15’ N, 20˚35’ E, 88 m a.s.l.), 25 km southeast of the city Zrenjanin in the Vojvodina region, northern Serbia. The mean annual air temperature and annual precipitation recorded at the nearby climate station of Zrenjanin is about 11.2˚C and 622 mm (with high interannual variability and a dry period during the summer months) (Hrnjak et al., 2014; Tošić et al., 2014). The studied section is situated at the edge of the smallest loess plateau in the Vojvodina, the Tamiš loess plateau. This slightly elevated geomorphological unit is located between the floodplain of the Tamiš River (on the northeast, east and southeast) and the palaeochannels Petra and Šozov on the west and northeast (Popov et al., 2012; Fig. 5.1). The Orlovat section is located at the southeast edge of this plateau and the profile was probably influenced by slope processes during its formation. The published luminescence chronology and absence of L1SS1 (units nomenclature according to pan-European loess stratigraphic model proposed by Marković et al., 2015) pedocomplex indicates a hiatus in sedimentation between around 40 and 13 ka (Marković et al., 2014a).

5.2.2. Sampling strategy

During April and May 2012, the Orlovat profile was carefully cleaned and sampled for sedimentological analyses, including rock magnetic properties (Marković et al., 2014a), sediment color (Lukić at al., 2014), granulometry and geochemistry. Samples for rock magnetic properties (Marković et al., 2014a), color of the sediment (Lukić et al., 2014) and grain-size analysis were taken in 5 cm resolution, while the geochemistry samples were taken in 10 cm resolution with 5-10 cm gaps around transition zones. Eight samples were collected for luminescence dating using metal tubes (Marković et al., 2014a; Timar-Gabor et al., 2015). Samples from the modern soil were taken only from the lowermost ca. 20 cm, because the uppermost ca. 40 cm of the soil have been influenced by human activities. This study reports results from grain-size, multi-elemental (XRF) and CaCO3 content analyses.

Fig. 5.1. The study area. (A) Map of the Vojvodina region with the geographical positions of the main loess sections (Markovićet al., 2014a, modified). (B) A geomorphological map of the Tamiš loess plateau surrounded by the Tamiš and Begej river valleys (Popov et al., 2012, modified).

5.3. Material and methods

5.3.1. Grain-size and geochemical analysis

All samples were air-dried, homogenised and sieved (<2 mm). Subsamples of 0.1–0.3 g fine- earth (< 2mm) were pre-treated with 0.70 ml of 20% hydrogen peroxide (H2O2) at 70 °C for 12 hours. This process is repeated until a bleaching of the sediment occurs (Allen and Thornley 2004), but not longer than three days. To keep particles dispersed, the samples were treated with 1.25 ml, 0.1 M sodiumpyrophosphate (Na4P2O7 * 10H2O) for 12 h (Pye and Blott, 2004). The particle size was measured by a LS 13320 Laser Diffraction Particle Size Analyser (Beckman Coulter), which divides the samples in 116 grain-size classes with a range of 0.04–2000 μm with an error of 2% (1σ). Each sample was measured four times in two different concentrations to increase the accuracy. Afterwards, all measurements with reliable obscuration were averaged. To calculate the grain-size distribution the Mie theory was used (Fluid RI: 1.33; Sample RI: 1.55; Imaginary RI: 0.1) (Özer et al. 2010; ISO 13320-1, 1999). The particle size fractions were defined by employing the ISO standard 14688 (2002), where clay is represented with particles smaller than 2 μm, fine silt from 2 to 6.3 μm, medium silt from 6.3 to 20 μm and coarse silt from 20 to 63 μm (Blott and Pye, 2012).

In past studies of Serbian LPS based on conclusions from studies that compared grain-size data using the sieve and pipette method with laser measurements (Antoine et al 2003, 2009; Konert and Vandenberghe, 1997), the clay fractions were proposed to be the fractions <4.8 μm (Antoine et al., 2009), <5 μm (Obreht et al., 2014) or <5.5 μm (Bokhorst et al., 2009) for laser grain-size measurements. Based on the results of the Orlovat loess sequence and the different characteristics of the measurement instruments, the clay fraction is defined to be <2 μm. However, also the particles <5 μm are reported for a better comparability with already published data.

For the present study, the element concentrations of 10 major elements and 14 trace elements were determined (Figs. 5.5 and 5.6, Supplementary Tabel 5.1). The bulk sediment samples were sieved down to 63 μm and dried at 105˚C for 12 h. An 8 g-quantity of the sieved material was mixed with 2 g Fluxana Cereox wax, homogenised and pressed to a pellet with a pressure of 19.2 MPa for 120 s. The measurements were conducted by means of a pre- calibrated method. Loess and palaeosol samples were analyzed in duplicate for major and

trace element abundances with polarization energy dispersive X-ray fluorescence (EDPXRF) using a SpectroXepos.

The CaCO3 – content was determined gas-volumetrically using a SCHEIBLERapparatus after dissolution of carbonates with HCl (10%) (Schaller, 2000; ISO 10693, 1995).

5.3.1.1. Ratios and chemical weathering indices

Chemical weathering indices are based on the concept on mineral alteration, where the selective removal of soluble and mobile elements from a profile section is compared to a relative enrichment of immobile and non-soluble elements (e.g. Fedo et al., 1995; Harnois, 1988; Kronberg and Nesbitt, 1981; Yang et al., 2004). Ratios of bulk elements and weathering indices have been successfully used for the reconstruction of palaeoenvironmental conditions of LPS (e.g. Jeong et al., 2008; Kels et al., 2014; Muhs et al., 2008; Varga et al., 2011; Yang et al., 2006). The chemical weathering indices used in this study are the Chemical Index of

Alteration (CIA= ((Al2O3/(Al2O3+Na2O+CaO*+K2O))*100) (Nesbitt and Young, 1982), the

Chemical Proxy of Alteration (CPA) = (Al2O3/(Al2O3+Na2O)) * 100 (Buggle et al., 2011)

(also known as CIW´ (Cullers, 2000)) and (CaO + Na2O + MgO)/TiO2 ratio (Yang et al., 2006).

The A-CN-K diagram (Al2O3-(Na2O+CaO*)-K2O diagram) (Nesbitt and Young, 1984) informs about weathering and sorting effects of aluminosilicates, as well as the initial composition of the unweathered material (e.g. Nesbitt and Young, 1989; Nesbitt et al., 1996).

5.4. Results

5.4.1. Grain-size distributions and their change with stratigraphy

The grain-size density distribution curves from the individual stratigraphic units of the Orlovat section are presented in Fig. 5.2, and the high resolution profile records of the individual grain-size fractions, GSI (Antoine et al., 2009) and U ratio (Vandenberghe and Pissart, 1993; Vandenberghe et al., 1998) are given in the Fig. 5.3.

Figure 5.2. Density distribution curve for the main stratigraphic units of the Orlovat section.

Figure 5.3. The grain-size proxies, U-ratio, GSI and CaCO3 content related to the pedostratigraphy. Ages shown in ka next to the sequence represent the results of luminescence dating (Marković et al., 2014a).

The grain-size <2 and <5 µm fractions have very similar distributions (see Fig. 5.3) and are therefore described together. From the bottom of the profile at 10 m to ~9.4 m, the relative contribution of clay fractions increases from ca. 12.3 to ca. 20.6% and the fine (<5 µm) fraction increases from 20.3 to 30.3%. Until 5.4 m the contribution of this fine fraction decreases almost linearly to 8.8% for clay and 14% for <5 µm fractions, and then increases almost to its maximum values at ca. 4.2 m (19.5% for clay and 28.6% for fine particles). From 4.2 to ca. 2.5 m a general decreasing trend can be observed. From here until the top of the section the data have no trend, with maximum values for this part of the section around 1.3 m (17.5% for clay fractions).

Also the fractions 2-22 and 5-22 µm have similar distributions and thereby only 2-22 µm will be described. From the bottom of the profile at 10 m to ca. 8.1 m the amount of the medium silt fractions decreases from ca. 39 to 27%. From 8.2 to 5.4 m data are similar with low fluctuations around 29%. From 5.4 to 5 m the medium silt fractions increase from 27% to 34.5%. Up to 4 m values are similar in the range between 33 and 35%. From 4 to 1.8 m the distribution of the medium silt fractions increases to ~42.7%. Above 1.8 m the medium silt particles have large fluctuations in their abundance to the top of the profile with the highest amount of 48.7% at 0.5 m depth.

The coarser silt particles (22-63µm) generally have low fluctuations within the Orlovat section ranging from 29-37.5%. From the base of the profile to 9.4 m, the contribution of the coarse silt decreases from 36.2 to 30.7%. From there to 7.4 m values are stable in the range of 30 to 32%. An increase of coarse silt from 7.4 to 5.5 m was interrupted by a decrease between 6.6 and 6.4 m. From ~5.5 to 4 m coarse silt decreases from 37.7 to 28.9% and then increases until 2.2 m to the maximum contribution of 37%. From 2.2 to 0.5 m the contribution shows a decreasing trend with large fluctuations.

The sand fractions range from 8 to 30%. From the bottom of the profile to 8 m, the sand fractions increase from 10 to 27.8%. From 8 to 5.4 m the sand fractions reach highest amount of 29.9% with no obvious trend. From 5.4 to 1.8 m, the sand fraction decreases. Above 1.8 m, the sand fractions slightly increase with large fluctuations and no trend can be observed.

Granulometry of the Orlovat section is depicted in Fig. 5.4, where fractions >44 μm of the Orlovat section are compared to the same fraction distribution of the Surduk section (Antoine et al., 2009) at the Srem loess plateau (south of the Fruška Gora Mountains, see Fig. 1) and 62

the Titel section (northwest of the confluence of the Danube and Tisa, Marković et al., 2008) at the Titel loess plateau. The Orlovat section contains higher concentrations of coarser fractions (>44 μm) in the palaeosol S1 than the Surduk and Titel sections. Contrarily, the Surduk and Titel sections contain coarser fractions within the loess. Furthermore, the palaeosol S1 at the Orlovat section is much ticker than at the other sections, while the thickness of the Last Glacial loess unit L1 is considerably lower at Orlovat.

Figure 5.4. Direct comparison between >44 μm fractions obtained from the Orlovat, Titel (Marković et al., 2008) and Surduk (Antoine et al., 2009). The profiles are plotted on their depth scales.

The Supplementary Figure 5.1 presents the normalised grain-size distribution of particles <63 μm. It can be observed that the clay fractions have a similar pattern as the bulk sediment grain-size distribution. The highest changes can be observed for the coarse silt fractions which are relatively stable in the bulk sediment distribution.

5.4.2. Geochemical analyses

5.4.2.1. Calcium carbonate

The amount of CaCO3 in bulk samples of the Ortlovat section is presented in Figs. 5.3 and

5.5-5.7. The CaCO3 content is high and varies from 9.2% to 31.8% (average 19.3%). From the bottom of the profile to 9.5 m the CaCO3 content increases to 31.8%. In the palaeosol S1 unit, the CaCO3 content gradually decreases from 9.5 to 6.5 m, with lowest values of 10.3%, while it increases again between 6.5 and 5.5 m. The increase in CaCO3 content continues until

5 m (highest values are ~22.5%). Above 5 m the CaCO3 content rapidly decreases and has stable values around 15% until 2.8 m. Between 2.8 and 2.2 m the CaCO3 content shows an abrupt increase up to ~26.5%. Constantly high values occur between 2.2 and 0.5 m, while the modern soil is highly depleted in CaCO3 (9.2%).

5.4.2.2. Major and trace elements

The concentration of major elements is presented in Fig. 5.5 and Supplementary Table 5.1. All the major element values are normalised and presented as percentage values. Despite the methodological approach using the sand-free samples, SiO2 is still dominant in all samples.

The SiO2 values vary between 51.5 and 63.1% (average 56.1%). Al2O3 reaches values between 12.8% and 15.42% (average 14.3%). CaO is also a major contributor to the sediments at Orlovat, and ranges from 7.8% to 20.4%, with an average of 14.4%. The FeO content varies from 5.1 to 6.1% (average 5.7%). The MgO contribution varies from 7.8 to 3.7% (average 4.6%) and its distribution does not show any correlation with other minerals.

Other elements are present in much smaller concentrations. K2O contributes with 2.00 to

2.73% (average 2.3%), Na2O with 0.99 to 1.49% (average 1.2%) and TiO2 is in the range of 0.90 to 1.09% (average 1%). MnO contributes with 0.10 to 0.14% (average 0.11%) to the

sediment. All major elements (except P2O5) show a clear shift of values between 2.8 m and 2.2 m.

Figure 5.5. The rock magnetic proxies (Marković et al., 2014a), major elements (normalised and presented in percentages values) and CaCO3 content related to the pedostratigraphy.

Trace element (Rb, Sr, Ba, Pb, Th, Zr, Nb, Y, V, Cr, Ni, Cu and Zn) contents are displayed in Fig. 5.6 and Supplementary Table 5.1. The distinct shift between 2.8 m and 2.2 m of all elements can also be observed in the trace element concentrations.

All trace and major elements except MgO, P2O5 and Sr have a similar pattern of distribution (except that CaO has the opposite trend). In general, the decreasing trend of values from the bottom of the profile to 9.5 m is followed by an increasing trend of values until 6.5 m. Between 6.5 m and ~3.2 m a general decrease of values is observed, whereas between ~3.2 and 2.6 m values show an increasing trend. An abrupt decrease in values between 2.6 and 2.2 m is observed, continued with gradual decrease to the top of the section.

MgO is characterised by an increase in values from the bottom to 9.8 m, an abrupt decrease can be observed until 9.4 m (from ~8-~4%). Above that up to 6.3 m values are stable around 4%. From 6.3 to 3.8 m values are increasing with highest peaks at 6.1, 5.4 and 4 m. Above values decrease until 2.8 m, followed by an increase until 2.2 m. From 2.2 m to the top of the profile values gradually decrease. Sr mirror the MgO with the opposite trend.

The P2O5 content shows almost constant values in a narrow range from 0.19 to 0.27%, except for the uppermost 1 m where the values rapidly increase to 0.75% (average 0.3%).

Figure 5.6. The rock magnetic proxies (Marković et al., 2014a), trace elements (presented in ppm values) and CaCO3 content related to the pedostratigraphy.

5.4.2.3. Elemental ratios and weathering indices

Elemental ratios are presented in Fig. 5.7. The Th/Ni ratio does not show any particular trend. The values are scattering, but an abrupt shift in values can be observed from 4.3 to 4 m. The Zr/Ni and Zr/Si ratios show no trend below 4.2 m. From 4.2 to 4 m, a rapid increase is observed for these ratios. Until 2.5 m both ratios show a decreasing trend. From 2.5 m to the top of the profile, the Zr/Ni ratio shows an increasing trend while the Zr/Si ratio has relatively stable values in a narrow range. The Al2O3/K2O ratio has a quite narrow range of values from the bottom of the profile to 5 m. From there to 2.8 m the values decrease. From 2.8 m to the top of the profile values are in a narrow range, and show a small increase from 1.5 to 0.6 m. Thereafter, values decrease in the modern soil. From the bottom of the profile to 2.5 m, the

SiO2/Al2O3 ratio shows values in a relatively narrow range with slightly little higher values from 7.5 to 5.5 m and 4.5 to 2.5 m. From 2.5 to 2.1 m, values rapidly decrease. From 2.1 m to the top of the profile a gradual decreasing trend can be observed.

The Ba/Sr and Rb/Sr ratios are similar and are therefore described together. Both ratios show a decrease in values from the bottom of the section to 9.5 m. From there to 8.5 m values increase. From 8.5 to 6.3 m depth, values are fluctuating without a trend. From 6.3 to 5.2 m values decrease again. Between 5.2 and 2.3 no trend can be observed. From 2.3 to 2.1 m values rapidly decrease, and from 2.1 to 1.6 m values are relatively stable. Above 1.6 m, values gradually increase until the top of the modern soil.

Weathering indices are presented in Fig. 5.7. The values of CIA and CPA are scattering and do not show any systematic enrichment in any particular litostratigraphic unit. A gradual decreasing trend characterised by a large fluctuation of values can be observed in the S1 palaeosol pedocomplex. The strong enrichment of values can be observed from 5.4 to 3.8 m, with an abrupt decrease of values at 4 m. Above, values decrease until 3.5 m. From there to 2.7 m no trend is observed and values are characterised by large fluctuations. From 2.7 m to the top of the profile values have an increasing trend with great fluctuations and an abrupt decrease at ~1.4 m.

The (CaO + Na2O + MgO)/TiO2 ratio increases from the bottom of the section until 9.5 m and then decreases up to 8.6 m. A weaker decreasing trend continues up to 6.2 m. From 6.2 to 2.3 m values do not show a clear trend, have small fluctuations and vary between 25 and 35. From 2.3 to 0.6 m values are generally high. In the uppermost 0.6 m values rapidly decrease to the minimum values at the top of the profile.

All ‘‘Na-type’’ weathering indices (CIA, Index B, CIW, PIA, and the CPA) show a very similar trend, therefore here only the CIA and the CPA are presented (Fig. 5.7). As the analyses were not done on CaCO3 free samples, the (CaO + Na2O + MgO)/TiO2 ratio is influenced by the CaCO3 content. Therefore is not reliable for weathering and will not be discussed here.

Figure 5.7. The rock magnetic proxies (Marković et al., 2014a) simple ratios, weathering indices and related CaCO3 content related to the pedostratigraphy. Ages shown in ka next to the sequence represent the results of luminescence dating (Marković et al., 2014a).

5.4.3. Stratigraphy and chronology

A detailed description of the Orlovat section and stratigraphic interpretation was presented by Marković et al. (2014a). A stratigraphic labeling scheme for the Vojvodina was established by Marković et al. (2015), where the loess and palaeosol stratigraphic units were designated with letters L (loess) and S (soil), and are numbered with increasing age.

It should be noted that the usual stratigraphic unit for MIS 3, a weak interstadial pedocomplex (L1SS1) in the Vojvodina region, is missing. The stratigraphic units are defined by Marković et al. (2014a) as follows, starting at the bottom of the profile: (I) L2 represents loess accumulated during MIS 6. (II) S1 is characterised by the interglacial palaeosol corresponding to MIS 5. This interglacial soil complex is divided into three different palaeosol formations. (III) L1 represents a loess unit mainly corresponding to the Last Glacial (mostly represented by loess matching roughly the time of MIS 2 and MIS 4, but also the Holocene loess in uppermost part). Finally, (IV) S0 represents the modern soil (Fig. 5.4, Table 5.1).

Table 5.1. Correlation of lithological units with Marine isotope stages (MIS) and depth.

Stratigraphic Sedimentary Depth units MIS units (m) L2 6 unit 1 10-9.5 unit 2 9.5-8.1 S2 5 unit 3 8.1-6.6 unit 4 6.6-5.5 unit 5 5.5-4.0 L1 4-2 unit 6 4.0-2.8 unit 7 2.8-2.2 L1 2-1 unit 8 2.2-0.6 S0 1 unit 9 0.6-0.0

According to this lithostratigraphy (Marković et al., 2014a) and data obtained in this study, we revise the stratigraphy from Marković et al. (2014a), and divide the profile into nine different stratigraphic units (Figs. 5.3 and 5.7, Table 5.1). The lowermost four units are divided according to the lithostratigraphy, where unit 1 is L2 loess, unit 2 is a strongly developed olive brown and blocky AB horizon (9.5 m-8.1 m depth), unit 3 represens a light olive brown palaeosol (8.1-6.6 m depth) and unit 4 includes three granular palaeosol horizons. Further division of the profile is based on different patterns in grain-size or element concentration observed in L1 loess, and presented as follow:

Unit 5 is associated with an increase of clay particles after a transition from the palaeosol to a pure loess unit at ca. 4 m depth. The constant increase of clay is characterizing this whole unit until a weak interstadial palaeosol.

Above the weak interstadial palaeosol, unit 6 extends up to a depth of 2.8 m. This unit is characterised by a decrease of the clay content. In this unit clay and sand do not seem to be complementary. Both clay and sand decrease in contribution while the silt fraction increases.

Unit 7 extends 0.6 m from 2.8 m to 2.2 m depth. In this unit the clay fractions are relatively stable. Although the grain-size does not show high fluctuations or a trend, an abrupt shift can be observed in the content of the most chemical elements.

Unit 8 is presenting a loess layer with two weakly developed palaeosols. A very slight increase of clay particles is observed in the weak palaeosols when compared to the overlying

loess layer, but generally this part of the profile has stable clay values. This unit shows a different grain-size pattern compared to all units below because the high complementary fluctuations have been observed in medium and coarse silt.

Unit 9 corresponds to the modern soil. This unit is characterised by a relatively high input of sand and coarse silt.

5.5. Discussion

5.5.1. Granulometry as an indicator of wind dynamics

Previous studies (Lukić et al., 2014; Markovć et al., 2014a) investigating the Orlovat section showed that this section was exposed to different sedimentological conditions during the past ~160 ka. The granulometry for the Orlovat section confirmed a unique sedimentology for the Carpathian Basin, due to a higher sand content in the soil and elevated clay content in the loess (Fig. 5.3).

The grain-size distribution curves of all stratigraphic units (Fig. 5.2) from the Orlovat section indicate aeolian sedimentation throughout. It can be concluded that redeposition did not play a major role in the sedimentation process, though a minor contribution of slope processes cannot be excluded. It is particularly interesting to investigate the input of the coarse fractions in the soil and the highest contribution of clay content in the loess.

To explain the cause(s) of a higher content of coarser fractions in the soils, two factors appear most likely here: (1) stronger wind activities and/or a (2) different/closer source of the sediment material during soil formation could cause this pattern. According to the lithostratigraphy and the datings, the S1 soil corresponds to MIS 5. No strong wind activities and high dust fluxes have been inferred from any other LPS in the Carpathian Basin for this time period. High sedimentation rates of loess are only characteristic for the glacial period (Bokhorst et al., 2011; Buggle et al., 2009; Fitzsimmons et al., 2012; Újvári et al., 2010, 2014b). However, during recent conditions, the southern Banat is under the influence of a strong southeast wind called Košava (Barbu et al., 2009; Unkašević, et al., 2007). It is possible that the Košava wind prevailed in the southern part of the Banat also during previous interglacials, and that it may have been strong enough to influence a large region.

An additional argument for the Košava wind as a dominant wind during the palaeosol formation at the Orlovat section is the presence of the Deliblatska peščara (Deliblato Sands) as a potential source area. The Deliblato Sands, a sand field situated southeast from the Orlovat section (Fig. 5.1), could have acted as a source of sand which has contributed to the modern soil and the S1 palaeosol at Orlovat. It was proposed that alluvial fans and overbank deposits of big lowland rivers are the main source of loess material in Carpathian Basin (e.g. Smalley and Leach, 1978; Smalley et al., 2009). However, also local rocks exposed in the surrounding mountains of specific loess sites may contribute to the aeolian dust input (e.g. Újvári et al. 2012). In case of the Orlovat section, the Deliblato Sands southeast of the section exposed to the strong southeast Košava wind could strongly contribute to the grain-size content. Therefore, we propose the Deliblato Sands to be an important source area of aeolian sediments during the MIS 5 interglacial in addition to the river systems.

This hypothesis can be seen as an indicator of a dominant southeast wind during MIS 5 in the South Banat region as a whole. Furthermore, this southeastern wind may have an influence on a larger region, because the higher contribution of coarser fractions in the S1 palaeosol is also observed at the Titel section (Marković et al., 2008), relative to the Surduk section (Fig. 5.4). The suggestion that the Košava could have influenced the sedimentation on the Titel loess plateau is in agreement with findings by Zeeden et al. (2007), who observed that the depressions on the Titel loess plateau have a preferential northwest-southeast orientation, probably influenced by aeolian deflation. Those morphological patterns indicate that the Košava wind may have (had) a stronger influence on the Carpathian Basin than previously known, because the direction of landforms are aligned in the direction of the Košava, and beyond these geomorphological properties high contents of coarse material is present in a large area.

Marković et al. (2008) pointed out that during the Last Glacial cycle, a higher deposition of coarser material occurred during the Early Pleniglacial in the south Carpathian Basin. Contrarily to that, the Orlovat section contains a relatively small contribution of coarser particles during the Early Pleniglacial compared to the other sections in the Carpathian Basin (Fig. 5.4). Bokhorst et al. (2011) proposed that beside reduced wind speed, a shift in the source area caused by a change in the wind direction may have influenced accumulation rates and contributions of coarser material. Generally it is proposed that the Carpathian Basin was under the influence of north and/or northwest winds during the Last Glacial (Bokhorst et al., 71

2011; Marković et al., 2008; Sebe et al., 2011). Therefore it is more likely that a higher contribution of finer and medium fractions at the Orlovat section witness the change of wind directions, and thus provenance, than wind speed. A shift from southeastern winds during MIS 5 to north/northwest winds during MIS 4 would imply a change in the source area. There are no source areas rich in coarse particles north of the Orlovat section, and therefore north/northwest winds would not be suitable for the accumulation of coarse material. Concluding, the most likely scenario would be that during MIS 5 strong southeast winds were dominant over the south Banat region, possibly having an influence on the wider region, while during MIS 4 north/northwest winds prevailed. During this period, the investigated section was most probably continuously exposed to far distance transport of particles.

Sedimentation rates during MIS 3 and MIS 2 were low or/and sediment formed within this period was eroded (see Sectiopn 5.5.4) and Orlovat may not present a continuous record. Therefore it is not possible to reconstruct aeolian detailed dynamics for this period from Orlovat. According to the luminescence ages (Timar-Gabor et al., 2015), the uppermost ~2 m (units 8 and 9) mostly present Holocene sediment. It may be expected that during the Holocene a higher sand abundance would be recorded at the Orlovat section as the southeast winds should characterize interglacials. Although there are some excursions of the higher sand content, they are well below the high values observed during the formation of the S1 palaeocomplex. However, comparing the Orlovat section to the other sections in the Carpathian Basin, the Holocene part is characterised by very high accumulation rates (in particular unit 8). Since no high accumulation was observed elsewhere, southeastern winds could explain these unusually high accumulation rates at Orlovat. A lower sand contribution in comparison to the S1 palaeosol may be explained by the additional new and closer source area which was not present during S1 palaeosol formation (see Section 5.5.4.). This new source area could dilute the contribution of sand originating from the Deliblato Sands with finer particles. Also, high accumulation rates induced by the strong wind activities could explain the delay in onset of the Holocene soil formation. Hence, it can be concluded that southeast winds prevailed during the Holocene.

It can be reliably asserted that southeastern winds prevailed during MIS 5 and the Holocene, but we are more careful with the statement that the termination of these wind occurred during MIS 4. It is known that during the Holocene until the 18th century (before it was anthropogenically forested) the Deliblato Sands represented an active dune field area 72

(Bukurov, 1982). It might be assumed that during MIS 5 this was also the case. However, it may be possible that during the Last Glacial, the Deliblato Sand was consolidated under a stable vegetation cover, and therefore, could not be the source area for the coarser particles. Even a higher precipitation during the interglacial may not have been a strong controlling factor of the plant cover in sandy areas, as it percolates fast. A higher evapotranspiration during interglacials is not favourable for a regular vegetation growth as well. A dune field with no vegetation cover would have been more active and would have possibly been located closer to the Tamiš loess plateau than today (see Fig. 5.1). During the Last Glacial, although the precipitation was lower, the lower evapotranspiration could provide a higher soil moisture favouring a vegetation cover on the dune field. Some studies have even proposed a pattern of “warm and dry interglacials” and “cold and humid glacials” as a potential regional phenomenon (Obreht et al., 2014; Zech et al., 2009; Zech et al., 2013). Therefore, the vegetation cover of the dune field during the glacial could have prevented further deflation of sand and its accumulation in the loess layers. In addition, very high sedimentation rates observed at the Titel loess plateau during glacial conditions (Fig. 5.4) could be better explained with southeast winds, since the Danube and Tisa rivers as the main source area are located south and east from the plateau. Similar conclusions can be applied to the eastern part of the Srem loess plateau, where the Danube as the main source area is located east of the plateau. Observed high sedimentation rates could be hardly explained by the north/northwest winds proposed by Sebe et al. (2011). It is possible that contrary to the north Carpathian Basin, where the north/northwest winds prevailed (Sebe et al., 2011), the south Carpathian Basin was permanently under the influence of southeastern winds. Besides, it can be confidently claimed that southeastern winds were prevailing at the south Banat region during the Last Interglacial, we here cannot prove or rule out their existence during glacials.

5.5.2. A grain-size perspective for the reconstruction of palaeoenvironmental conditions

As already shown in study of Obreht et al. (2014), clay fractions do not necessarily represent a sensitive palaeoclimate proxy. It was noted earlier that the usual grain-size patterns, in particular in the clay content, do not necessarily document enhanced rates of pedogenesis during the Holocene and MIS 5 (Stevens et al., 2011), as higher sediment accumulation rates during glacial periods may mask evidence of pedogenesis such as clay mineral formation. The

higher clay content in loess units can be a product of continuous weathering during low accumulation rates. The abundance of sand particles in clay-poor layers and vice versa speaks in favour of this hypothesis.

In the Carpathian Basin, the magnetic susceptibility is generally enriched in palaeosols due to a higher amount of ferrimagnetic minerals formed during pedogenesis and clearly reflects changes in pedo- and lithostratigraphy (e.g. Buggle et al., 2009, 2014; Liu et al., 2013; Marković et al., 2009). However, Marković et al. (2014a) observed irregularities in the magnetic susceptibility at the Orlovat section associated with a gradual transition between the bottom of the S1 and the L1. This transition of magnetic susceptibility is comparable to the contribution of clay and sand fractions in the S1 palaeosol, and can be explained by the grain- size distribution in the S1 palaeosol. Sand fractions usually have a lower magnetic susceptibility than other grain-size fractions due to the diamagnetic properties of quartz. The correlation between an increase of sand and a decrease of magnetic susceptibility can indicate that the sediment was influenced by an input of sandy material during the soil formation. Higher values in magnetic susceptibility should be expected if the sand was present before the soil formation started (e.g. Jary and Ciszek, 2013). It is very likely that units 3 and 4 represent a syngenetic soil that formed while coarser material accumulated and did not overprint the already formed palaeosol. The high sedimentation rates characterised by coarser fractions could partly mask pedogenetic features, such as the clay translocation and the magnetic susceptibility, but they may not completely mask the formation of soil. We can conclude that at the Orlovat section the granulometry is not only dependent on palaeoenvironmental and palaeoclimatic conditions, but also on the textural distribution of the accumulating material, which might have different source areas.

The unusually high sedimentation rates during the humid periods have also been observed in the Chinese Loess Plateau (e.g. Stevens et al., 2008; Stevens and Lu, 2009). At the Orlovat section, the high sedimentation rates of coarser particles did not completely mask the formation of soil(s). Also, low accumulation rates observed in the upper part of unit 5 and unit 6 (characterised by a higher clay contribution) did not enable soil formation. This might indicate that soil formation during warm and humid periods (integlacials) cannot easily be masked by high sedimentation rates. Furthermore, low sedimentation rates cannot (always) enable soil formation during cold and dry period such as glacials since precipitation is an important factor. According to δ13C values from the Carpathian Basin (Hatté et al., 2013; 74

Obreht et al., 2014; Schatz et al., 2011; Zech et al., 2013), Mediterranean-like climate with precipitation in spring and autumn was dominant during the last interglacial, as the precipitation during the warmest period will increase the abundance of C4 vegetation (Yang and Ding, 2006). Although summer precipitation was low, the high productivity of biomass was enabled over the year (Obreht et al., 2014). Lukić et al. (2014) presented that the Melanization Index (the color measurement proxy indicating an accumulation of humus and humic substances in the soil) shows much higher values in the S1 palaeosol than in loess units with rich clay fractions. We propose the annual duration of vegetation cover and biomass productivity during the year (probably related to the length of the growth period) as an important factor for the formation of humus. This also shows that soil formation during periods with high biomass production such as MIS 5 could not be easily masked by increased accumulation rates. During glacial conditions, the annual production of biomass was probably much lower due to a shorter growth period.

5.5.3. Geochemical characteristics (simple ratios and weathering indices)

From major elements, Al2O3 and CaO are highly influenced by the CaCO3 content throughout the whole section. A similar pattern can be observed for Na2O and K2O. MgO does not show any correlation to grain-size and CaCO3 content. However, just the simple comparison of element contents can be biased due to a systematic enrichment or dilution of carbonate and quartz minerals. On the other hand, element ratios are usually a reliable proxy for the geochemical characterization of the sediment. The simple ratios Ba/Sr and Rb/Sr are assumed to indicate weathering and have been widely used in the studies of the Carpathian Basin (Bokhorst et al., 2009; Buggle et al., 2011; Újvári et al., 2008, 2014a; Varga et al., 2011) (Fig.

5.7). Buggle et al. (2011) suggested that these ratios are influenced by the CaCO3 content. Schatz et al. (2014) indicated that such a correlation has not been observed at the Tokaj sequence (Hungary). At the Orlovat section, Ba/Sr and Rb/Sr ratios are not strongly influenced by the CaCO3 content. A reason for this may be that Sr is highly correlated to MgO at the Orlovat section. This should not be considered as a general rule, as in other studies from the Carpathian Basin (e.g. Bokhorst et al., 2009; Buggle et al., 2008; Varga et al.,

2011) it seems that the Sr content is controlled by the distribution of both, MaO and CaCO3.

Although weathering indices highly fluctuate, the general trends clearly show similarities to grain-size data. Decrease in finer fractions indicates a decrease of Al2O3, while increase in

coarser particles indicates an increase of Na2O and vice versa. The weak palaeosol at the top of unit 5 is characterised by low values of CIA and CPA, probably due to a high input of relatively coarse particles. One should be careful with interpretations of weathering indices in loess–palaeosol units, as these can be highly influenced by grain-size changes.

Unit 7 shows unsystematic and large fluctuations in the weathering indices, though grain-size distributions are relatively stable. Considering the luminescence ages, this may confirm that this layer contains erosional surfaces and relocated sediment. In contrast to the units below, units 8 and 9 provide CPA and CIA values that are characteristic for the Carpathian Basin region, having lower values in loess than in the modern soil. Some similarities can be observed between weathering indices and grain-size distributions, but they do not strongly correlate in these units. Therefore, the interpretation of weathering based on elemental ratios and weathering indices may be biased at the Orlovat section, as the ratio values are highly influenced by the grain-size distribution, especially in units 1–6.

5.5.4. Potential changes in source material

The provenance of coarser material and sand in the S1 palaeosol unit was discussed in Section 5.5.1. The Deliblato Sands are probably the source of coarser and sand fractions, whereas transport mechanism is most probably the Košava wind. However, provenance of fine and medium fractions cannot be concluded based on the dominant wind directions.

Although the elemental ratios and weathering indices do not reliably represent the weathering processes at the Orlovat section, some ratios, which are not influenced by weathering and pedogenesis, may be used to investigate the provenance of the sediment. The elements Th and Zr are enriched in felsic rather than in mafic rocks, because they are highly incompatible during most igneous melting and fractionation processes (Taylor and McLennan, 1985; McLennan, 2001). For example, Zr/Ni ratios are useful in some cases, as Zr is enriched in coarser fractions, whereas Ni is generally preserving a signature of the provenance (Újvári et al., 2014a). Also the Zr/Si ratio may reflect changes in the parent material. Th/Ni is a good indicator of igneous chemical differentiation processes since Th is incompatible, whereas Ni is a compatible element within the igneous systems (McLennan et al., 1993). The SiO2/Al2O3 ratio can represent a change in parent material, but also a change in variations of sedimentary

sorting linked to the distance of transportation (Hao et al., 2010). The Al2O3/K2O ratio is influenced by weathering, but changes in the ratio can show a dominance of K-feldspar contribution, as can be achieved by the A-CN-K diagram.

All studies performed in the Carpathian Basin (Schatz et al., 2014; Újvári et al., 2008) and the Lower Danube Basin (Buggle et al., 2008; 2011) have shown a common pattern of weathering presented within the A-CN-K diagram (Fig. 5.8). All samples from these studies plot on a line parallel to the A–CN join. This pattern presents a distribution of material with different extents of chemical weathering, resulting in a predominant removal of silicatic Ca and Na due to the greater destruction of plagioclase and a relatively slow removal rate of K from K- feldspar. At the Orlovat section samples are also plotted on the line parallel to the A–CN join. However, samples are not plotted within one line, as it was observed in studies before, but have three different patterns of distribution. Samples of units 1-5 vary on the line that has a similar distribution as in the previous studies. The distribution line of samples of unit 6 are slightly dislocated toward the K apex, and the distribution line of samples from units 7-9 are even more dislocated toward the K apex (Fig. 5.8). Although there are three different patterns, the distribution of values within each pattern varies exclusively on the line parallel to the A- CN join. Hence, these patterns are not influenced by differences in K-feldspar weathering, as the weathering line shows a clear destruction predominantly of plagioclase, but rather a change in parent material K-feldspar content. The samples of units 1-5 are similar to results from previous studies and contain less K-feldspar than the other units at this section. Unit 6 presents a transition zone from an old to a new provenance of material enriched in K-feldspar, whereas units 7-8 are completely exposed to a new source. This can be also observed using the Al2O3/K2O ratio (Fig. 5.8).

Since units 1-5 have similar weathering pattern as other sections in the Carpathian Basin, it may be assumed that the provenance for these units can be found in the Danube and/or Tisa River sediments. Buggle et al. (2008) and Újvári et al. (2008) concluded that the geochemical results cannot prove an accurate provenance, but rather can rule out some areas as source area. Accordingly, neither the Danube nor Tisa could be more precisely favoured or ruled out as a source area here. However, it can be claimed that sediments from one or both of these rivers were the main sediment source area. Even a contribution of particles from the Deliblato Sands should not have a very different signal, as these particles primary originate from Danube

sediment. Differences are expected to be limited to grain size and sorting during aeolian transport.

A gradual change in source of material within unit 6 can be best explained by a gradual approach of new source material. Only rivers could be a source area that can approach to the Orlovat section gradually. Since the geochemical imprint of the new source differs from the Danube and Tisa River signals, the only possible source could be the relatively small Tamiš and Begej Rivers. However, the catchment areas of these rivers are geochemically unexplored and therefore this hypothesis cannot be confirmed or ruled out. However, relatively low and stable (do not exceed 50 ppm; Fig. 5.6) Ni contents exclude all areas with mafic and ultramafic rocks as source area. Therefore, the Tamiš and/or Begej Rivers as a new source area and the southwest Carpathian Mountains as a provenance area seem to explain observed changes well.

Figure 5.8. A–CN–K diagram.

The enhanced fluvial dynamics of a palaeo-Tamiš river could explain the hiatus observed at the Orlovat section. Fluvial activities probably eroded sediment accumulated after ~40 ka. Studies from the south Carpathian Basin showed a relatively strong developed interstadial palaeosol (L1SS1) at ~40 ka (Fuchs et al., 2008; Schmidt et al., 2010; Stevens et al., 2011). Since a palaeosol did not develop/preserve, erosion before ~40 ka seems a plausible explanation.

According to the luminescence ages, units 7-9 mostly correspond to the Holocene. These units are characterised by K-feldspar rich material, probably originating from the Carpathian Mountains. However, not only the K-mineral abundance suggests a change of source material, but the SiO2/Al2O3 ratio (Fig. 5.7) gives evidence of a new pattern in sorting of grain-size, which is linked to different transport distances. A higher SiO2/Al2O3 ratio in the units 8 and 9 confirms a change in the distance of the source area. Additionally, small rivers establish a new grain-size pattern in units 8 and 9 where complementary fluctuations can be observed in medium and coarse silt, whereas the fluctuations of medium silt are mostly influenced by sand in the units below. Therefore, it can be assumed that sedimentation during late MIS 2 and the Holocene was mainly influenced by smaller rivers (according to the current situation influenced by the Tamiš River). Contrary, during MIS 4 the sedimentation was mainly connected to long-distance transport.

5.6. Conclusions

Observations from previous studies suggested that the Orlovat section was exposed to different sedimentological conditions as compared to other sections studied in the Carpathian Basin; we here confirm this observation. The grain-size and geochemical data enable us to provide a better insight into sedimentological conditions. Despite the slope position, it seems that slope-related processes did not play a major role. The sedimentological conditions are probably related to different geomorphological features in the vicinity of the section. Differences in the dominance of wind direction played a major role for the sediment accumulation. We assume that the high sand input of the S1 palaeosol and also elevated input into the modern soil is strongly influenced by the Deliblato Sands as source are. The Košava wind with a southeast-northwest wind direction is proposed as a transport mechanism and the most dominant wind for the South Banat region. The source material for the accumulation of loess has changed a few times during the last ~160 ka at the Orlovat section, and it is assumed

that the source and its distance from the Orlovat section triggered different palaeosignals and influenced the sediment accumulation rates. Therefore, we would like to highlight the importance of understanding changes of source area and material, as well as factors controlling loess accumulation, such as wind directions and sediment supply.

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

Igor Obreht1, Christian Zeeden1, Ulrich Hambach2,3, Daniel Veres4,5, Slobodan B. Marković3, Janina Bösken1, Zorica Svirčev3, Nikola Bačević6, Milivoj B. Gavrilov3, Frank Lehmkuhl1

1Department of Geography, RWTH Aachen University, Templergraben 55, 52056, Aachen, Germany

2Chair of Geomorphology & BayCEER, University of Bayreuth, 94450 Bayreuth, Germany

3Laboratory for Paleoenvironmental Reconstruction, Faculty of Sciences, University of Novi Sad, Trg Dositeja Obradovića 2, 21000 Novi Sad, Serbia

4Institute of Speleology, Romanian Academy, Clinicilor 5, 400006 Cluj-Napoca, Romania

5Interdisciplinary Research Institute on Bio-Nano-Science of Babes-Bolyai University, Treboniu Laurean 42, 400271 Cluj-Napoca, Romania

6Department of Geography, Faculty of Natural Sciences and Mathematics, University of Kosovska Mitrovica, Lole Ribara 29, 38220, Kosovska Mitrovica, Serbia

Published in Scientific Reports 6, 36334

Abstract

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.

6.1. Introduction

Knowledge of past climate variability based on the study of palaeoclimate archives may help clarify past climatic forcing mechanisms and help predict the extent of future climate change (Ganopolski et al., 2016; Yin and Berger, 2015). In Southeastern Europe loess-palaeosol sequences are one of the most important and often the only available terrestrial archives of Quaternary palaeoclimate dynamics (Marković et al., 2015; Sprafke and Obreht, 2016). Loess-palaeosol sequences from the Middle and Lower Danube Basins (Fig. 6.1) have been the focus of recent research on the Quaternary in Europe (Buggle et al., 2014, 2013; Fitzsimmons et al., 2012; Marković et al., 2014a, 2011, 2008; Necula et al., 2015; Újvári et al., 2008), and as a result, the understanding of the Late Quaternary climate and environmental conditions in the region have been fundamentally improved. Although the studies of loess sequences in Southeastern Europe have provided local and regional information of past environments, they have not focused on identifying large scale forcing mechanisms and the climatic conditions responsible for loess-palaeosol sequences formation,

particularly whether or not a Mediterranean climatic influence may have existed in the region. One of the main reasons for this lack of data is that the Middle and Lower Danube Basins are influenced by three different climatic systems: Atlantic, continental and Mediterranean. It is challenging to distinguish which of these climate systems may have dominated during the recent past. Also, the loess-palaeosol sequences as palaeoenvironmental records in the Middle and Lower Danube Basins are generally limited in terms of their climate sensitivity due to the overall effect of prolonged dryness in the region (Buggle et al., 2013; Marković et al., 2015, 2008, 2007), which is an additional limiting factor in establishing the relative influence of associated atmospheric systems.

The Central Balkans (Fig. 6.1) are situated in a transition area between the temperate- continental climate in the north and the Mediterranean climate in the south. The area is currently more influenced by the Mediterranean climate than are the Middle and Lower Danube Basins (Kostić and Protić, 2000). Although palaeoclimate records from the Central Balkans have been poorly investigated (Basarin et al., 2011; Kostić and Protić, 2000; Obreht et al., 2014), this area may be key to understanding past changes between these two climate zones. To address this, we conducted a high-resolution multi-proxy investigation of the Stalać section in the Central Balkans (Serbia; Fig. 6.1). For this site, we examined grain-size, environmental magnetism, and other spectrophotometric and geochemical proxies to reconstruct past climatic and environmental dynamics for the past 350,000 years. Grain-size composition is one of the most frequently used palaeoenvironmental proxies of loess sequences to infer changes in aeolian dynamics, sources of loess and pedogenesis (Luehmann et al., 2013; Schaetzl and Attig, 2013; Újvári et al., 2016; Vandenberghe, 2013). Based on the principle of post-depositional formation of ultrafine magnetic particles during pedogenesis, environmental magnetism has also been successfully applied as a palaeoclimatic proxy in Eurasian loess (Buggle et al., 2014; Hao et al., 2012; Marković et al., 2011). Color variations in loess research have also been used as a proxy for variations in mineral concentrations and as an indicator of past pedogenesis (Buggle et al., 2014; Zeeden et al., 2017). Finally, the composition of geochemical elements has been successfully applied to establish temporal weathering intensity in Southeastern Europe (Buggle et al., 2013; Újvári et al., 2014a) and also used to establish the provenance of loess in same region (Buggle et al., 2008; Újvári et al., 2008).

Figure 6.1. Map of the Balkan Peninsula, Middle and Lower Danube Basins, showing key loess-palaeosol sequences (yellow rectangles; Stalać (this study), Ruma (Vandenberghe et al., 2014), Titel (Bokhorst et al., 2011), Batajnica (Buggle et al., 2014, 2013, 2009), Orlovat (Marković et al., 2014a; Obreht et al., 2015), Belotinac (Basarin et al., 2011; Obreht et al., 2014), Mircea Voda (Buggle et al., 2014, 2013, 2009)) and lacustrine records (red rectangles; Ohrid (Sadori et al., 2016), Prespa (Leng et al., 2013) and Tenaghi Philippon (Tzedakis et al., 2006)) discussed in this paper. The white dashed line represents the current northern limit of the Mediterranean climate. The map was generated using ArcGIS 10.2.2 (http://www.esri.com/software/arcgis/arcgis-for-desktop/free-trial).

The main goal of this study is to better understand the degree to which Mediterranean and continental climate zones have influenced this part of Southeastern Europe, by comparing loess deposits in the Central Balkans and palaeo-archives from the surrounding areas (the continental Middle Danube Basin, the South Balkans and the Mediterranean itself). Differences observed in the characteristics of loess sediments in these regions may help

identify past spatial variations in the dominant atmospheric systems over Southeastern Europe. The interpretation of the relationships between the Mediterranean and continental climate zones would be a significant step toward a wider understanding of the past atmospheric circulation patterns over continental Europe at large. Past research has suggested that Southeastern Europe has been progressively influenced by drier continental climate over time (Buggle et al., 2009). Thus, past climatic intervals, proposed as the best analogues to modern conditions according to astronomically driven orbital changes on global scale (e.g. Marine Isotope Stages (MIS) 11 and 19 (Yin and Berger, 2015)) may not strictly be representative in this region, since those periods experienced more humid conditions. In such a region with no comparable analogues, research on the interaction between large scale climate systems and their feedback mechanisms is important to the understanding of future climate change. Besides addressing regional teleconnections, we also evaluate the influence of global climate changes on the continental part of Southeastern Europe.

6.2. Results 6.2.1. Stratigraphy and chronology

Figure 6.2 presents the Stalać composite stratigraphic profile built from five sampled subsections; for details see the Supplementary material Chapter 6. The labeling of the stratigraphic units follows the established scheme for the Danube loess stratigraphy (Marković et al., 2015). The age model is based on the correlation of odd-numbered MIS to phases of soil formation; details are presented in the Supplementary material Chapter 6.

The composite profile (Fig. 6.2) shows the sedimentological characteristics of loess-palaeosol sequences accumulated during the past ~350,000 years. The stratigraphic profile commences with a L4 loess unit (uncovered only 0.4 m). In the top of this unit is a strongly developed, brown-red Cambisol with vertic characteristics and a thickness of ca. 1.25 m (S3). This palaeosol is overlain by a layer of typical loess, ranging in thickness from 1.65 to 3 m (L3). From 3.0 to 3.45 m, another Cambisol with vertic characteristics is exposed (S2), but this one is less strongly developed than the S3 palaeosol. The S2 palaeosol is followed by a 5.4 m thick, pale yellow loess layer (L2). A volcanic tephra layer is intercalated within the L2 loess (4.85-4.9 m). The L2 loess ends in a 0.85 m thick Kastanozem-type palaeosol (S1), overlain by a brown-yellow loess layer, approximately 1.30 m thick (L1LL2). This layer terminates in a 1.1 m thick palaeosol (L1SS1SSS2), covered by a 0.7 m thick but darker loess layer 85

(L1SS1LLL1). Microscopic investigation of this layer shows high abundance of well- preserved glass shards, hinting at the presence of a cryptotephra horizon within this layer. Above, another palaeosol horizon (12.85-13.65 m) is developed (L1SS1SSS1). A brownish loess layer is exposed from 13.65-14.7 m (L1LL1), on top of which is developed the modern topsoil (S0). The upper part of the modern soil has been heavily influenced by modern agriculture, therefore only the lowermost 30 cm have been sampled. Material in Profile 1 was deposited on the different geomorphological conditions than in profiles 2-5, and therefore it is challenging to compare the sedimentation rates from profile 1 to the rest; further details are available in Supplementary material Chapter 6.

Figure 6.2. Clay fractions, U-ratio, χ, Ni contribution, L* and a* values (see Methods) related to pedostratigraphy of the composite profile from the Stalać section.

6.2.2. Particle size properties and environmental magnetism

The entire section is comprised of loess and is silt-dominated, particularly coarse silt (20-63 µm sized particles). When compared to other Eurasian loess-palaeosol sequences the Stalać section has a high sand (>63 µm sized particles) content (Supplementary Fig. S6.5). The

grain-size distribution (see Supplementary Fig. S6.9) suggests that only the L1SS1LLL1 layer is not unaltered loess; instead, it may have been affected by special sedimentological conditions, showing a unique peak in sand contribution. Generally, all palaeosols are comparatively enriched in clay (<2 µm particles) with highest values in S3 (maximum of 28 %); the various intervening loess units contain less clay (minimum of 4.7 %; Fig 6.2). The U ratio (coarse/fine silt (16-44/5.5-16 µm); Fig. 6.2) has been used as a proxy for wind strength (Vandenberghe et al., 2014). The mass specific magnetic susceptibility (χ; Fig. 6.2) varies between 82.1*10-8 and 371.8*10-8 m3/kg (average 189*10-8 m3/kg) almost an order of magnitude larger than for other loess deposits in the Danube Basin, Central Asia and China (Buggle et al., 2014). Maximum χ values occur in the L2 tephra layer, whereas high values occur in the younger parts of L3 and L1. Minimum χ values occur in the basal L4 unit. The L2 (excluding the tephra layer) also has low values of χ. The frequency dependent magnetic susceptibility (χfd) varies between 3.1 to 7.3 % (average 4.6 %). This proxy follows the lithostratigraphy (Supplementary Fig. 6.6), having higher values in palaeosols (maximum in S3) and lower values in the intervening loess units (minimum in L2) but not reaching the average level (higher of 10 %) of interglacial palaeosols in common loess-palaeosol sequences (Buggle et al., 2014).

6.2.3. Bulk sediment geochemistry

The sediment is dominated by SiO2 although with a large range, oscillating between 39.87 and 66.4 % (average 59.57 %); these values are typical for loess deposits (Buggle et al., 2008;

Újvári et al., 2008). Also Al2O3 (average 15.95 %), CaO (in a wide range from 0.81-36.36 %; average 9.22 %), FeO (average 6.34 %) and MgO (average 3.76 %) are strongly represented in the sediments, and this, too, is similar to other European loess-palaeosol sequences (Buggle et al., 2008; Újvári et al., 2008). All other elements comprise less than 3 %. A unique characteristic of the Stalać loess is the high concentration of Ni and Cr. Ni values range from 46.5 to 172.6 ppm (average 110.3 ppm) and Cr ranges from 99.4 to 310.6 ppm (average 210.6 ppm). Generally, Ni and Cr concentrations exhibit wide variations between and within different stratigraphic units. High values are recorded in L1 and L3, whereas other units have lower concentrations (Fig. 6.2). Supplementary Table 6.1 presents Ni and Cr values for alluvium from the nearby rivers. The Zapadna Morava alluvium has the highest concentrations, while the Južna Morava alluvium contains lower amounts of Cr and Ni.

6.2.4. Tephrochronology

Based on visual field observations, as well as trends in sedimentological data, two main tephra layers have been identified for the studied time-interval (Fig. 6.3 and Supplementary Fig. 6.7). Whereas glass shards from the visible tephra within L2 are completely altered and thus unsuitable for geochemical tracing of the volcanic source, the chemical investigation of glass shards from upper tephra in L1SS1LLL1 layer is presented in Supplementary Table 6.3.

With average SiO2 content of 59.25 wt%, associated with 18.14 wt% Al2O3, 7.34 wt% K2O,

5.39 wt% Na2O, 2.94 wt% FeO and 1.84 wt% CaO, the major oxide data confirm that this perialkaline trachitic cryptotephra layer is yet another occurrence of the regionally widespread Campanian Ignimbrite/Y-5 tephra (De Vivo et al., 2001; Fitzsimmons et al., 2013; Marti et al., 2016; Veres et al., 2013) (Supplementary Fig. 6.9 and Supplementary Table 6.3).

6.2.5. Spectrophotometric results

The spectrophotometric data from the Stalać composite profile is presented in Fig. 6.2 and Supplementary Fig. 6.6. Lightness (L*; see Methods) tends to indicate the production of biomass during the sediment formation. L* values (ranging from 52.66 to 75.13) are higher in loess and lower in palaeosol units. However, loess layers L1LL1, L1SS1LLL1 and L1LL2 are darker (lower L* values) than loess units L2 and L3. Redness (a*; red-green scale) correlates well with lithostratigraphy, having high values in palaeosols and low in loess units (except in L1SS1LLL1). Differences in colour between different palaeosols are clearly visible. The highest a* values occur in the S3 palaeosol (up to 6.82) and are somewhat lower in S2 (up to 5.48), while the values in S1, L1SS1SSS2 and L1SS1SSS1 do not exceed 5. The b* (blue- yellow) has its highest values in S3 (maximum 24.09) and S2 palaeosols, but lowest in S1, L1SS1SSS2 and L1SS1SSS1 palaeosols (minimum value 16.91). The modern soil has very low L* and high a* and b* values.

6.3. Discussion

To understand the significance of the presented data, it is important to realize that besides several available lacustrine records from the southern Balkan Peninsula (Tenaghi Philippon (Tzedakis et al., 2006), Ioannina (Tzedakis et al., 2002), Kopais (Tzedakis, 1999), Ohrid (Sadori et al., 2016) and Prespa (Leng et al., 2013)), other high resolution records in the

interior are still missing in the Balkan region (Fig. 6.1). Comparing the climate evolution from the interior of the Balkans with records from Mediterranean climate areas, such as the Tenaghi Philippon lacustrine record (Tzedakis et al., 2006), can provide useful insights into the Mediterranean climate circulation patterns over the Balkans. Moreover, understanding the relations between Stalać and Lake Ohrid (Sadori et al., 2016) is of special importance. Because of its altitude of 693 m a.s.l. and its considerable distance from the sea (Fig. 6.1), this lake is currently under a modified Mediterranean climatic influence and is more sensitive to continental climate than are other lacustrine records in the Balkans. Finally, understanding of the relations between Stalać and the well-studied loess-palaeosol sequences in the Middle Danube Basin improves the understanding of the interplay between the Mediterranean and continental climates in Southeastern Europe during the Quaternary.

The geochemical characteristics of the Stalać loess-palaeosol sequence are characterised by high concentrations of Ni and Cr (Fig. 6.3 and Supplementary Fig. 6.11), values that unequivocally indicate surrounding river valleys (Južna, Zapadna and Velika Morava Rivers) as the source areas (for more details see Supplementary material Chapter 6). This is also supported by the higher contribution of the coarser grain-size particles, when compared to middle Danube basin (Bokhorst et al., 2011; Obreht et al., 2015; Vandenberghe et al., 2014). Such geochemical and textural characteristics suggest limited or completely suspended contribution of the particles from the Danube catchment area, which is the main source area for the terrestrial sediments in the Middle and Lower Danube Basins. Despite the relative vicinity of the Danube River (Fig. 6.1), the absence (or very limited presence) of particles from its catchment area at the Stalać section indicates that the Central Balkans were not strongly influenced by the atmospheric circulation patterns and winds from the north. A sharp climatic border/transition between the southern limit of Middle Danube Basin and the area north of the Stalać section can explain the observed pattern.

Figure 6.3. Direct comparison between the benthic δ18O LR04 stack (Lisiecki and Raymo, 2005), MEDSTACK planktic δ18O data (Wang et al., 2010), arboreal pollen from Tenaghi Philippon (Tzedakis et al., 2006), arboreal pollen (Pinus, Juniperus and Betula are excluded) from the Lake Ohrid core (Sadori et al., 2016), U-ratio and <2 µm fractions from the Stalać section (plotted versus age, abscissa), U-ratio and <2 µm fractions from the Ruma section (Vandenberghe et al., 2014) (the only section with existing grain-size record spanning the last three glacial-interglacial cycle in Middle Danube Basin; plotted vs. depth). Note the differences between scales on the plots presenting Stalać and Ruma grain-size data. Tephra layers are marked with yellow lines; “CI” refers to the Campanian Ignimbrite/Y-5 tephra layer and “L2” refers to the L2 tephra layer.

Also, essential differences between these regions are observed in the palaeosol development within their respective sections, especially during MIS 9 and 7. For those interglacials, at the Stalać section the intercalated palaeosols are classified as Cambisols. These palaeosols are reddish in color (high a* values) and have high clay contents (Fig. 6.2). These data indicate that the Mediterranean climate influence was strong in this area during these interglacials (e.g. see references Buggle et al. (2013) and Buggle et al. (2014)). Conversely, palaeosols in the Middle Danube Basin indicate the overall dominance of steppe or forest-steppe environments during the past ~350,000 years (Marković et al., 2012b), pointing to the influence of the continental climate during these interglacials (Buggle et al., 2013). Although MIS 7 and 9 in the Central Balkans were characterised by generally more precipitation than the Middle Danube Basin, summers were warmer and much drier. Accordingly, during MIS 7 and 9, the Balkan Peninsula was under the strong influence of a sub-tropical anticyclone belt during the warm seasons, creating low precipitation during summer in the Mediterranean region, and thus, a Mediterranean-like climate dominated. Contrary, in the Middle Danube Basin these interglacials had relatively similar conditions to the present conditions, i.e., being under continental climate (cold winters and warm summers, with a precipitation maximum during the beginning of the warm period). Observed similarities in the Stalać data and the pollen data from Tenaghi Philippon (Tzedakis et al., 2006), Lake Ohrid (Sadori et al., 2016) and planktonic 18O data from the Mediterranean (MEDSTACK) (Wang et al., 2010) during MIS 9 - MIS 6 (see Supplementary material Chapter 6) generally indicate similar climate conditions across the Balkans (Fig. 6.3).

However, although the Mediterranean-like climate was dominant during this time interval over the Central Balkans, differences in clay content and redness between palaeosol S3 (related to MIS 9) and S2 (related to MIS 7) clearly indicate the weakening of Mediterranean climate influence over time (Fig. 6.4); this trend is not observed in the other geoarchives of Mediterranean region (Fig. 6.3). Figure 6.4 presents a succession of palaeosol types from Cambisols towards steppe-like palaeosols, with a concomitant decrease in <5µm particle peak values in the related palaeosols at the Stalać section (Central Balkans) and the Batajnica-Stari Slankamen (spliced) section9 (Middle Danube Basin). Although data from the Stalać section do not span the same time interval, a progressive diminution of Mediterranean climate influence can be observed within the Southeastern Europe. This progressive evolution of continental climate (aridization and winter cooling) in the Middle and Lower Danube Basins

since the Early Pleistocene was proposed to be caused by the uplift of the Alps, Carpathians and Dinarides (Buggle et al., 2013). For the Central Balkans, only the uplift of the Dinarides is likely to have played a major role in this climate shift. However, considering that the termination of the Mediterranean climate is more pronounced in the Central Balkans over the past 350,000 years, it is unlikely that the uplift of the Dinarides could have weakened the Mediterranean influence in the observed proportions. The present study demonstrates that the progressive evolution of the continental climate in the Middle Danube Basin and the progressive decrease of the Mediterranean climate influence in the Central Balkans are probably connected, although this trend is more expressed in the Central Balkans over the past 350,000 years. Additionally, a similar trend of increased continentality over time is observed in the Azov Sea region (Liang et al., 2016). Thus the observed trend is not only a regional trend for the Middle and Lower Danube Basins, but may be a supra-regional climatic trend. A more complete understanding of mechanisms behind this pattern requires more long- term high-resolution data.

Figure 6.4. Progressive termination of Mediterranean influence over Southeastern Europe, as indicated by the succession of palaeosol types and <5µm particles peak values in related palaeosols at (a) the Stalać section (Central Balkans) and (b) the Batajnica-Stari Slankamen spliced section (Middle Danube Basin; Buggle et al., 2013). 92

All the long palaeoclimate records for the Balkans (Sadori et al., 2016; Tzedakis et al., 2006) (including Stalać) specify MIS 6 as the most intense glacial period with the most adverse conditions, and with a maximum of glacier expansion (Hughes et al., 2006, 2007, 2011). MIS 6 is also considered to represent the most intense glaciation in the rest of Europe (Ehlers and Gibbard, 2007). Geochemical tracing of palaeo-river systems, based on the contributions of Ni and Cr (for more details see Supplementary material Chapter 6), shows that the L2 loess (equivalent to MIS 6) is the only such layer formed mainly by particles from the Južna Morava River valley (Fig. 6.2 and Supplementary Fig. 6.11). This suggests that dry conditions over the Zapadna Morava River catchment strongly reduced its discharge, whereas the discharge from the Južna Morava catchment was not so dramatically limited by the general dryness of this period (see Supplementary material Chapter 6). This is a clear indication of dry conditions over the entire interior of the Balkans, including high mountainous areas, suggesting that precipitation could not reach the interiors of the Balkans. Possible reasons for this may be (1) specially cold conditions that would have increase aridity, (2) the maximal expansion of glaciers in the Balkan mountain ranges (Hughes et al., 2011, 2007, 2006) causing a high-pressure barrier for the moist air from the south, and (3) increasing distance from the sea, the major source of humidity, due to the low sea level (Rohling et al., 2012) in this time interval.

Our study suggests a shift in the climatic and environmental conditions over the Central Balkans since MIS 5. The palaeosol S1 (equivalent to MIS 5) shows remarkably different features as compared to the older palaeosols at the Stalać section, indicating a limited Mediterranean climate influence. Similarities in the genesis of the S1 (MIS 5), L1SS1SSS2 and L1SS1SSS1 (MIS 3) pedocomplexes (Fig. 6.2) indicate similar conditions during the last interglacial and MIS 3 interstadials at the Stalać section. A high abundance of fine particles and low L* values from L1LL2 and L1LL1 loess layers (Fig. 6.2 and Supplementary Fig. 6.12) indicate that the transition from the last interglacial to the early last glacial was not sharp, and relatively mild conditions prevailed during the early and late last glaciation (Fig. 6.2). Most of the last glacial cycle at the Stalać section was relatively humid and mild compared to previous glacials. An exception is the L1SS1LLL1 layer, where volcanic glass shards were found (Supplementary Fig. 6.8), clearly showing that this layer represents a short accumulation event of volcanic ash and aeolian silt. The chemical composition of glass shards from this tephra layer (Supplementary Fig. 6.9 and Supplementary Table 6.3) suggests that

this tephra layer is another occurrence of the widespread Campanian Ignimbrite/Y-5 tephra, a crucially important marker horizon found in many locations throughout the Mediterranean, Balkans, and further east (Fitzsimmons et al., 2013; Veres et al., 2013). Originating in the Campi Flegrei volcanic field of Italy and dated elsewhere (40Ar/39Ar) to 39,280 ± 110 years BP (De Vivo et al., 2001), this is the first report of this tephra find within loess records in southeastern Europe outside the Lower Danube area (Fitzsimmons et al., 2013; Veres et al., 2013). The vegetation cover was probably significantly degraded due to the local impact of the ashfall (Fitzsimmons et al., 2013), enabling the higher availability and dynamics of coarser grain-size particles (Figs. 6.2 and 6.3). The timing of the Campanian Ignimbrite/Y-5 tephra eruption (De Vivo et al., 2001; Fitzsimmons et al., 2013; Marti et al., 2016; Veres et al., 2013) also closely matches the timing of Heinrich event 4, which may have resulted in a delay in vegetation recovery and generally colder climatic conditions even several centuries after the eruption. The high sand content indicates unfavourable environmental conditions during the short accumulation period of this layer (ca. 0.70 m) even in the upper part of this layer, where the proxies used for tephra determination (including microscopy) show a limited/suppressed contribution of the volcanic ash (see Supplementary material Chapter 6 and Supplementary Fig. 6.7). It can be concluded that Heinrich event 4 was linked to specially expressed stadial conditions in Southeastern Europe.

It has to be emphasised that studies on glaciers from the Balkans(Hughes et al., 2006, 2007, 2011) support rather mild last glacial conditions. Those studies showed that maximum glacier expansion in the last 350,000 years was during MIS 6 (Hughes et al., 2011, 2007, 2006), while the glacier expansions during MIS 10 and MIS 8 were rather similar to those during MIS 6 (especially over the Dinarides (Hughes et al., 2011)). However, MIS 5-2 were characterised by significantly less extensive and smaller glaciers (Hughes et al., 2011, 2007, 2006). The limited extension of glaciers during the Late Pleistocene, including the period of the last glacial maximum, over the Balkans supports either cold and very dry or, alternatively, generally milder conditions and less expressed glacial-interglacial climate variability. Data from our study support the second scenario with generally milder glacial conditions and/or warmer summer temperatures (higher abundance of fine particles, lower values of the U-ratio and lower L* relative to previous glaciations; Fig 6.3 and Supplementary Fig. 6.12). Similar suggestions have been made from the deep-sea temperature reconstructions based on independent sea level models from the Mediterranean Sea (Rohling et al., 2014), proposing

that the last glacial may have been relatively mild (Rohling et al., 2014). The milder last glacial conditions were caused by a feedback triggered by the shift from Mediterranean-like climate towards continental climate in MIS 5. An intensification of continental climate conditions over Southeastern Europe shifted the subtropical anticyclone to the south, causing an increase in precipitation during summer, and a decrease of precipitation during winter without significant change in the overall temperature relative to previous interglacials. This shift towards continental climate significantly reduced the extent of glaciers over the Balkans, also during the cold stages after MIS 5. The limited presence of glaciers during the last glacial cycle in the Balkans’ mountains enabled penetration of warm and moist air from the south into the interior of Balkans, leading to the mild glacial conditions in the interior of the Balkans as recorded at Stalać. The Greenland ice-sheet extent during the studied period may had an influence on the stronger continental climate over Central Balkans. A simulation of the Greenland ice-sheet extent during last ~350,000 years (de Boer et al., 2014a, 2014b) indicates that only the last interglacial (MIS 5e) was characterised by a large-scale melting of this ice- sheet (Supplementary Fig. 6.12). This may have established different atmospheric circulations over Europe at that time with a stronger influence of anticyclones from continental Europe (Siberian high) over the Balkans due to the weakening of Greenland high pressure. However, at this stage the inferred pattern is indicative only. If valid, it suggests that the melting of the Greenland ice-sheet can significantly change the future climate in the interior of the Balkans and intensity of the continental climate influence in Middle Danube Basin.

In summary, this study provides the first comprehensive palaeoclimate reconstruction of the Central Balkans for the past ~ 350,000 years based on the terrestrial sediment proxies. Data show that the Central Balkans experienced different palaeoenvironmental and palaeoclimatic conditions than the Middle and Lower Danube Basins did. The latter were much more strongly influenced by continental climate. Our findings suggest that a sharp climatic transition zone may have existed between these regions. A Mediterranean-like climate prevailed from MIS 10 to MIS 6 in the Central Balkans. However, our data also point to a general trend of progressively weaker Mediterranean climate influence within the entire Southeastern Europe (Fig. 6.4). An abrupt shift in general climate conditions over the interior of the Balkans to more continental conditions from MIS 5 to present is evident in the data derived from the section, possibly connected to the Greenland ice-sheet retreat during MIS 5e. The prevalence of the continental climate influence over the Central Balkans led to a change

in precipitation maxima from winter to warmer periods of the year, resulting in a significant reduction of glaciers in the Dinarides. The absence of large glaciers enabled the penetration of relatively warm and humid air from the south and southwest into the interior Balkans, leading to higher humidity during the last glacial, when compared to previous glaciations.

From a wider perspective, this study has established the spatial and temporal borders of the Mediterranean climate influence and its interaction with the continental climate influence over Southeastern Europe. However, the palaeoclimatic reconstruction of the Middle Danube Basin remains primarily based on environmental magnetic properties of loess-palaeosol sequences, especially for the sequences preserving older records than the last glacial. Specific physical properties and changes in sediment provenance at the Stalać section make it impossible to reconstruct the palaeoclimate based on environmental magnetism (for more details see the Supplementary material Chapter 6). A better understanding of the past climatic relations between these two regions requires more studies based on grain-size and colour analyses from long loess sections in the Middle Danube Basin, but also investigations of the older parts of the Stalać section.

Additionally, our study paves the way for new discussions, studies and hypotheses. For example, the observed different climate conditions in the Central Balkans and Middle Danube Basin indicates a narrow climatic boundary that might explain the role of the Balkan Peninsula as a Quaternary floristic refugium (Tzedakis et al., 2002). Such a climatic boundary is thought to have protected the Balkan Peninsula from the adverse climatic conditions in the North. Particularly, enhanced humidity during the last glacial explains why the Balkan Peninsula was an important floristic refugium during this period (Tzedakis et al., 2002) and highlights its importance as a European biodiversity hotspot. Our study also has the potential to improve our knowledge concerning the dispersal of anatomically modern humans from Africa to Europe, because the Balkans have been suggested as one of the main migration corridors. It questions the assumption that the appearance of the anatomically modern humans in the Balkans was determined by the unfavourable climate conditions before ~45 ka (Benazzi et al., 2011; Higham et al., 2011; Trinkaus et al., 2003). We emphasize that even though the last glacial was mainly characterised by relatively mild glacial conditions in the Central Balkans, the L1SS1LLL1 layer records the strong impact of the distal ashfall after the Campi Flegrei super-eruption (De Vivo et al., 2001; Marti et al., 2016)and the Heinrich event 4, indicating a drastic shift towards harsh environmental conditions in this area. Although the 96

timing and role of Campanian Ignimbrite/Y-5 ashfall, combined with a sudden climatic change and its interaction between the demise of the Neanderthals and their replacement by anatomically modern humans still remains a matter of hypothesis (Fitzsimmons et al., 2013; Lowe et al., 2012), our data suggest a strong impact of the Campanian Ignimbrite and Heinrich event 4 on this region. This supports the assumption that the Campanian Ignimbrite ashfall and a sudden climatic change during the Heinrich event 4 might have played an important role in the demise of Neanderthals, so that the anatomically modern humans could get a permanent hold in Europe only when the Neanderthal populations started to fade.

6.4. METHODS

Grain-size

Subsamples of 0.1–0.3 g fine-earth (< 2mm in diameter) were pre-treated with 0.70 ml of

30 % hydrogen peroxide (H2O2) at 70 °C for 12 hours. This process was repeated until a bleaching of the sediment occurs (Allen and Thornley, 2004), but not longer than three days. To keep particles dispersed, the samples were treated with 1.25 ml, 0.1 M sodiumpyrophosphate (Na4P2O7 * 10H2O) for 12 h (Pye and Blott, 2004). Particle size characteristics were measured with a LS 13320 Laser Diffraction Particle Size Analyser (Beckman Coulter). To calculate the grain-size distribution the Mie theory was used (Fluid RI: 1.33; Sample RI: 1.55; Imaginary RI: 0.1) (ISO International Standard 13320, 2009; Schulte et al., 2016). Clay is represented with particles smaller than 2 μm, fine silt from 2 to 6.2 μm, medium silt from 6.2 to 20 μm, coarse silt from 20 to 63 μm and sand higher than 63 µm (Blott and Pye, 2012).

Environmental magnetism

The volumetric magnetic susceptibility was measured at frequencies of 300 and 3000 Hz in a static field of 300 mA/m using a Magnon International VSFM. Data were corrected for drift and for the effect of sampling boxes (weak diamagnetism), and normalised to density. Hence, magnetic susceptibility is given as mass specific susceptibility (χ) in m3/kg. The frequency dependence was calculated as χfd = (χlf – χhf) / χlf *100 [%] (Buggle et al., 2014; Forster et al., 1994).

Bulk sediment geochemistry

All bulk sediment samples were sieved down to 63 μm and dried at 105˚C for 12 hours. An 8 g quantity of the sieved material was mixed with 2 g Fluxana Cereox wax, homogenised and pressed to a pellet with a pressure of 19.2 MPa for 120 seconds. The measurements were conducted by means of a pre-calibrated method. Samples were analysed for major and trace element abundances with polarization energy dispersive X-ray fluorescence (EDPXRF) using a SpectroXepos.

Glass shard chemical analysis

The geochemical analyses of glass shards were performed on a sample from layer L1SS1LLL1 with the highest Cl values. The sediment sample was sieved and glass shards were isolated through density separation, mounted in epoxy resin, ground and polished in preparation for microprobe analysis. Measurements were made using single-grain, wavelength-dispersive electron microprobe analysis at the Bayerisches GeoInstitut on a Jeol JXA8200 microprobe employing an accelerating voltage of 15 keV. A 6 nA beam current and defocused beam were used. Order of measuring elements (first to last): Na, Si, K, Ca, Fe, Mg, Al, P, Ti, Mn, Cl-. Peak counting times were 10 s for Na, 30 s for Si, Al, K, Ca, Fe and Mg, 40 s for Ti and Mn, and 60 s for P. Precision is estimated at <1–6% (2σ) and 10–25% (2σ) for major and minor element concentrations respectively. Analytical settings are presented in the Supplementary Table 6.4.

Spectrophotometric analysis

A Konica Minolta CM-5 spectrophotometer was used to determine the colour of dried and homogenised sediment samples by detecting the diffused reflected light under standardised observation conditions (2° Standard Observer, Illuminant C). Colour spectra were obtained in the visible range (360 to 740 nm), in 10 nm increments, and these data were then converted into the Munsell colour system and the CIELAB Colour Space (L*a*b*) using the Konica Minolta SpectraMagic NX software. The resultant values indicate the extinction of light on a scale from L* 0 (absolute black) to L* 100 (absolute white), and express colour as chromaticity coordinates on red-green (a*) and blue-yellow (b*) scales.

Correlative age-model

The age model is based on simple correlation of palaeosols to odd MIS (Lisiecki and Raymo, 2005) (S0 – MIS 1; L1SS1 and L1SS2 – MIS 3, S1 – MIS 5, S2 – MIS 7 and S3 – MIS 9) and loess layers to even MIS (L1LL1 – MIS 2, L1LL3 – MIS 4, L2 – MIS 6, L3 – MIS 8 and upper part of L4 – late MIS 10), as established for the Middle Danube Basin (Basarin et al., 2014; Buggle et al., 2009; Marković et al., 2012b). For more information see Supplementary material Chapter 6.

7. Shift of large-scale atmospheric systems over Europe during late MIS 3 and implications for Modern Human dispersal

Igor Obreht1, Ulrich Hambach2,3, Daniel Veres4,5, Christian Zeeden1, Janina Bösken1, Thomas Stevens6, Slobodan B. Marković3, Nicole Klasen7, Dominik Brill7, Christoph Burow7, Frank Lehmkuhl1

1Department of Geography, RWTH Aachen University, Templergraben 55, 52056, Aachen, Germany

2 BayCEER & Chair of Geomorphology, University of Bayreuth, 94450 Bayreuth, Germany

3Laboratory for Paleoenvironmental Reconstruction, Faculty of Sciences, University of Novi Sad, Trg Dositeja Obradovića 2, 21000 Novi Sad, Serbia

4Romanian Academy, Institute of Speleology, Clinicilor 5, 400006 Cluj-Napoca, Romania

5Interdisciplinary Research Institute on Bio-Nano-Science of Babes-Bolyai University, Treboniu Laurean 42, 400271 Cluj-Napoca, Romania

6Department of Earth Sciences, Uppsala University, Villavägen 16, 75236 Uppsala, Sweden

7Institute of Geography, University of Cologne, Albertus-Magnus-Platz, 50923 Cologne, Germany

Published in Scientific Reports 7, 5848

100

Abstract

Understanding the past dynamics of large-scale atmospheric systems is crucial for our knowledge of the palaeoclimate conditions in Europe. Southeastern Europe currently lies at the border between Atlantic, Mediterranean, and continental climate zones. Past changes in the relative influence of associated atmospheric systems must have been recorded in the region’s palaeoarchives. By comparing high-resolution grain-size, environmental magnetic and geochemical data from two loess-palaeosol sequences in the Lower Danube Basin with other Eurasian palaeorecords, we reconstructed past climatic patterns over Southeastern Europe and the related interaction of the prevailing large-scale circulation modes over Europe, especially during late Marine Isotope Stage 3 (40,000-27,000 years ago). We demonstrate that during this time interval, the intensification of the Siberian High had a crucial influence on European climate causing the more continental conditions over major parts of Europe, and a southwards shift of the Westerlies. Such a climatic and environmental change, combined with the Campanian Ignimbrite/Y-5 volcanic eruption, may have driven the Anatomically Modern Human dispersal towards Central and Western Europe, pointing to a corridor over the Eastern European Plain as an important pathway in their dispersal.

7.1. Introduction Knowledge of the past climatic interaction of large-scale atmospheric systems controlling hydroclimate variability is of wide importance because it improves our understanding of global climate evolution and may help constrain predictions of future climate changes. Nevertheless, it is still challenging to reconstruct the dynamics and impact of past large-scale atmospheric circulation systems. This is especially true over the European continent because Europe receives competing influences of the Atlantic, Mediterranean and continental climates, and palaeoclimate proxies preserving the imprint of such climatic interaction signals are not intensively studied or more often, difficult to quantify. Probably the best region to study past climate conditions related to the coherence of those climatic regimes is Southeastern Europe (Fig. 7.1), where these climate influences are interacting (Obreht et al., 2016; Panagiotopoulos et al., 2014; Stevens et al., 2011). Therefore, past changes in the strength and dynamics of such circulation modes are expected to be recorded in palaeoclimate archives from this region. Additionally, one of the oldest known fossils of Anatomically Modern Humans (AMH) in Europe were found in this area in the Peştera cu Oase cave (Western Romania; Fig. 7.1) and dated to ~40,000 years BP (Trinkaus et al., 2003). As such, 101

this region may be regarded as a key area for understanding the relationship between past environmental changes in this area and the dispersal of AMH throughout Europe during the late Marine Isotope Stage (MIS) 3.

In Southeastern Europe, loess-palaeosol sequences are one of the most important and usually the only available terrestrial archives of Quaternary palaeoclimate dynamics (Buggle et al., 2013; Marković et al., 2015; Obreht et al., 2015, 2016; Stevens et al., 2011; Zeeden et al., 2016). Although loess from Southeastern Europe has been in the focus of recent research, information about interactions among large-scale atmospheric systems that played a significant role in loess formation are scarce (Obreht et al., 2016). In our study, we reconstructed climatic conditions in the Lower Danube region during the past ~50,000 years using high-resolution grain-size, environmental magnetic and geochemical data supplemented by luminescence dating. Grain-size distributions reflect changes in aeolian dynamics, sources of aeolian dust and pedogenesis (Obreht et al., 2016; Újvári et al., 2016; Vandenberghe, 2013), whereas environmental magnetism indicates the post-depositional formation of ultrafine magnetic particles during in-situ weathering and pedogenesis, both linked to variations in soil humidity (Buggle et al., 2014; Zeeden et al., in press). Geochemical characteristics of the sediment can indicate the provenance or potential changes in the source area, and also give information on the weathering intensity. Based on the comparison between palaeoclimatic datasets from different regions in Europe we reconstruct the temporal and spatial interactions of Atlantic and continental climatic systems over Europe during the studied time period. Moreover, we evaluate the relations of inferred past atmospheric systems dynamics and their possible influence on the AMH dispersal throughout Europe.

102

Figure 7.1. Map of the Southeastern Europe showing key loess-palaeosol sequences (Urluia (this study), Vlasca (this study), Rasova (Zeeden et al., in press), Titel (Basarin et al., 2014; Bokhorst et al., 2011), Batajnica (Buggle et al., 2014, 2013), Orlovat (Marković et al., 2014; Obreht et al., 2015), Stalać (Obreht et al., 2016), Dunaszekcső (Újvári et al., 2016)) and lacustrine records (Lake Prespa (Panagiotopoulos et al., 2014) and Tenaghi Philippon (Tzedakis et al., 2006)) discussed in this paper. The map was generated using ArcGIS 10.2.2 (http://www.esri.com/software/arcgis).

7.2. Regional settings and study sites

Geomorphologically, Geomorphologically, Southeastern Europe is a diverse region, mainly mountainous throughout the Balkan area (Dinarides, Rhodope, and Balkan Mountains), whereas to the North the Carpathians separate two large lowland basins, the Middle Danube (Carpathian) Basin in the West and the Lower Danube (Walachian) Basin in the East, coinciding also with the westernmost extent of the Eurasian steppe belt (Fig. 7.1). This area 103

likely experienced major changes in the relative influence of large-scale atmospheric systems during the Middle and Late Pleistocene because of such geographical and geomorphological conditions (Buggle et al., 2009, 2013; Obreht et al., 2016). It is suggested that during the last glacial cycle the outermost extent of the Mediterranean climate regime was mainly limited to the Balkan Peninsula (Obreht et al., 2016). Conversely, the Middle and Lower Danube Basins were mostly under the influence of Atlantic and continental climates. The continental climate is characterised by expressed seasonality and moderate precipitation (concentrated mostly in the warmer months), while the Atlantic climate influence is related to the Westerlies. Similar to present-day conditions, the Lower Danube Basin was permanently under stronger continental climatic conditions during the Middle and Late Quaternary, when compared to the Middle Danube Basin (Buggle et al., 2013). Hence, a more detailed understanding of differences in the palaeoclimate evolution of the Middle and Lower Danube Basins provide information on temporal dynamics and spatial interaction of continental and Atlantic climates. Loess sequences from the Middle Danube Basin were studied intensively over the past decades (Marković et al., 2008, 2015; Stevens et al., 2011). However, besides notable advances in luminescence dating (Timar et al., 2010; Timar-Gabor et al., 2011), the Lower Danube Basin lacks high-resolution proxy data for the last glacial cycle. We performed high- resolution sediment analyses on the Urluia and Vlasca loess sections from the Lower Danube Basin, Romania (Fig. 7.1), for the past ~50,000 years. The sections were sampled in 2 cm increments for sedimentological, petrophysical and geochemical analyses. In this study, sediment fine fractions (<5 µm), the frequency dependent magnetic susceptibility (χfd) and weathering indices derived from geochemical analyses are used as an indicator of increased pedogenesis and weathering intensity, while the U-ratio (coarse/fine silt (16-44/5.5-16 µm)) (Vandenberghe, 2013) is taken as an indicator of the wind strength.

7.3. Results

The most relevant proxy data of the Urluia and Vlasca sections are presented in Figs. 7.2-7.3, while the stratigraphy and more detailed descriptions are given in the Supplementary Material Chapter 7. Both sections are comprised of loess and are silt-dominated, with a particular domination of coarse silt. The Vlasca section is characterised by a higher contribution of coarse particles with the average mean grain-size being 44.5 µm (Fig. S7.5) and an average sand content of 22.0 %. The Urluia section is composed of slightly finer particles (average mean grain-size 37.9 µm and average sand content of 17.9 %; Fig. S7.4). The contribution of 104

fine particles (<5 µm) varies from 9.3 to 23.2 % (average 13.8 %) at Vlasca and from 11.7 to 30.2 % (average 16.1 %) at Urluia (Fig. 7.2). The U-ratio is in the range from 2.1–3.9 at Vlasca and varies between 1.7 and 3.3 at Urluia (Fig. 7.2). The lowermost part of the Vlasca section (from the bottom to 10.1 m depth) is represented by a sandy layer (up to 58.3 % sand) and it is probably related to Danube River activities, and therefore is excluded from the palaeoclimate reconstruction.

Figure 7.2. Direct comparison between proxies from the Urluia and Vlasca sections represented by U-ration, fine particles (<5 µm) and χfd (values are normalised), and their comparison with a stacked climatic record from northern China (CHILOMOS (Yang and Ding, 2014)) and δ18O record from Dim Cave (Ünal-İmer et al., 2015) over the past 55,000 years. The straight yellow line represents the timing of the Campanian Ignimbrite/Y-5 tephra deposition. Black arrows indicate a trend of general continentalization.

105

Figure 7.3. Direct comparison between χfd record from Urluia (green line) and Vlasca (blue line), χfd from Titel loess-palaeosol sequence (Basarin et al., 2014) (orange line; Middle Danube Basin) and their comparison with the mean annual lake surface temperature (LST; red line) of the Black Sea (Wegwerth et al., 2015) over 25,000-50,000 years ago.

106

The mass specific magnetic susceptibility (χ) from the Vlasca section varies between 25.3*10-8 and 111.6*10-8 m3/kg (average 39.7*10-8m3/kg) and for the Urluia section from 17.5*10-8 to 86.6*10-8 m3/kg (average 28.4*10-8m3/kg; Fig. S7.6). The frequency dependent magnetic susceptibility (χfd) at Vlasca varies in a wide range between 1.8 and 11.3% (average 4.7%) and at Urluia between 2.5 and 11.7% (average 4.2%; Fig. S7.4-S7.6).

It is important to note an offset in χ (Fig. S7.6) and χfd (Figs. S7.4 and S7.6) at the Urluia section between the oriented samples that preserved their original moisture content and structure, and the sediment samples that have been dried, gently homogenised and compressed during the preparation process. Although all samples were density normalised, there is a difference between the density normalised susceptibilities of loess that preserved its original structure and moisture in oriented samples and such of dried and compressed loess. Although density normalised values are higher for non-oriented samples (Figs. S7.4 and S7.6), the general trends between the records are the same. Therefore, the values are normalised to a common base when presenting the whole profile of Urluia (Fig. 7.2) to avoid issues in displaying data.

Detailed geochemical investigations were performed at Urluia (every fifth sample was measured allowing the resolution of 10 cm) and Vlasca (with higher resolution of 2 cm increments, but with a focus on the samples between 41,500 and 5,500 years ago) sections. Both sections show a generally similar geochemical composition. The sediments are dominated by SiO2, oscillating between 50.7 and 66.2% (average 58.0%) at the Urluia section and between 50.7 and 66.2% (average 57.8%) at the Vlasca section. CaO (for Urluia between

4.8 and 24.4%, average 14.7%; for Vlasca between 9.2–19.8%, average 13.8%), Al2O3 (average for Urluia 13.5% and for Vlasca 13.3%), FeO (average for Urluia 5.0% and for

Vlasca 4.7%), MgO (average for Urluia 4.0% and for Vlasca 4.6%), K2O (average for Urluia

2.3% and for Vlasca 2.2%), Na2O (average for Urluia 1.2% and for Vlasca 1.3%) and TiO2 (average for Urluia and Vlasca is 1.0%) are also major contributors to the geochemical composition of sediments (Fig. S7.9), while all other elements comprise less than 1%. Geochemical investigations are also often used for better understanding of weathering intensity (Buggle et al., 2011; Krauß et al., 2016; Yang et al., 2004). Numerous ratios between soluble and mobile elements and immobile and non-soluble elements are widely used to give an insight into the weathering intensity of loess. Those ratios rely on the selective removal of soluble and mobile elements from a weathering profile compared to the relative enrichment of

107

immobile and non-soluble elements(Buggle et al., 2011; Yang et al., 2004). Among those, weathering indices such as the Chemical Index of Alteration (Nesbitt and Young, 1982)

(CIA= (Al2O3/(Al2O3+Na2O+CaO*+K2O))*100; CaO* is silicatic CaO) and the Chemical

Proxy of Alteration (Buggle et al., 2011) (CPA=(Al2O3/(Al2O3+Na2O))*100) are commonly used and widely accepted as reliable weathering indices in most environments (Fig. 7.4). At the Urluia section, the CIA is in the range between 58.4 and 65.7 and CPA between 85.3 and 90.2, while at the Vlasca section CIA varies between 62.5 and 69.3 and CPA between 83.4 and 89.0 (Fig. 7.4). Fluctuations of CIA and CPA show similar behavior at both sections (Fig. 7.4).

Figure 7.4. Weathering indices (CIA and CPA) of the Urluia and Vlasca sections.

108

Scanning electron microscope images of the glass shards (and also other magmatic and detrital aeolian grains) from a tephra layer at the Vlasca section are presented in Fig. S7.3, while Table S7.2 shows the geochemical composition of tephra glass shards. According to the geochemistry of the glass shards, this layer is unambiguously related to the Campanian Ignimbrite/Y-5 tephra.

7.3.1. Luminescence dating

Seven samples from the Urluia section were luminescence dated using the post infrared infrared protocol at 290 °C (Thiel et al., 2011). The samples show bright signals and all aliquots (10-24 per sample) passed the SAR rejection criteria. The prior IR stimulation temperature test of sample C-L3702 shows a plateau for temperatures between 50°C and 170°C (Fig. S7.1). Dose recovery tests (DRTs) are within 10% of unity for all samples (Fig. S7.1). Residual doses are < 10 Gy and were subtracted for DRTs, but not for equivalent dose

(De) measurements. Fading measurements show variable results with a mean g2days = - 0.9±0.7 %. De distributions show low relative standard errors <6%. Moreover, overdispersion values calculated by the central age model (CAM) (Galbraith et al., 1999) are small (<5 %), only sample C-L3715 shows a higher value of 9.4±1.6 %. α-efficiency measurements determined a mean a-value of 0.136±0.02. Ages and Des increase with depth from 21±1.6 ka (84.3±4.3 Gy) to 54.2±4.1 ka (258.2±13.3 Gy). Dose rates range from 4.0±0.2 Gy/ka to 5.3±0.3 Gy/ka. A summary of all relevant luminescence data is given in Table S7.1 and Fig. S7.2.

7.4. Discussion

The chronologies of both sections are primary based and linked to the occurrence of the Campanian Ignimbrite/Y-5 tephra dated to 39,930 ± 100 years BP (Veres et al., 2013). This layer serves as an excellent chronological marker horizon for loess deposits in the Lower Danube area. The age models used here are based on the ages obtained by luminescence dating and correlative techniques. Detailed information on establishing the age models are presented in the Supplementary Meterial Chapter 7.

Geochemical compositions of the Vlasca and Urluia sections are similar to other Danubean loess-palaeosol sequences (Buggle et al., 2008; Obreht et al.; 2016; Újvári et al., 2008), and the geochemical results from both sections do not indicate any major change in provenance

109

and source area, especially between 50,000 and 30,000 years ago (Fig. S7.10). The grain-size distributions from both sections are characterised by a higher contribution of coarser particles for loess sediment (Figs. 7.2, S7.4 and S7.5). The Vlasca section is situated nowadays in the immediate vicinity of the Danube River (Fig. 1) and it was likely also during the past 50,000 years under a strong influence of short-distance transported material, which is supported by the coarse particle contribution (Fig. S7.5). The Urluia section is more distant from the Danube (Fig. 7.1), and was subjected to receiving more distant material than the Vlasca section, although generally coarse particles still indicate a proximity to a source area (Fig. S7.4). Similar patterns in grain-size distribution between Vlasca and Urluia indicate that the recorded signal reflects a consistent regional pattern (Fig. 7.2). The grain-size distribution from both sections shows a higher contribution of fine particles before the Campanian Ignimbrite/Y-5 tephra deposition, indicating moderate wind dynamics from ~50,000-40,000 years ago. Upon the Campanian Ignimbrite tephra deposition, an increase in particles size and U-ratio values are recorded. Such a grain-size distribution indicates a clear trend towards stronger wind intensity and drier and probably colder climatic conditions during late MIS 3 (40,000-27,000 years ago) (Fig. 7.2). Low fine fraction contribution and high U-ratio values during MIS 2 indicate pronounced arid and cold conditions (Fig. 7.2).

Overall, both sections show changes in soil humidity indicated by variations in χfd that reflect a pacing similar to Dansgaard-Oeschger millennial-scale past climate variability (Fig. S7.6). However, due to the limitations in dating techniques in loess research, it is still not possible to reliably match fluctuations in loess records to such short climatic events. A comprehensive discussion on this issue is presented in Supplementary Material Chapter 7. Moreover, the χfd record from Vlasca seems to reflect changes in humidity with higher amplitude than Urluia (Figs. 7.2 and 3). This may be due to the local influence of the Danube River at Vlasca, whereas Urluia reflects the more regional humidity pattern. Although χfd shows differences in the amplitudes, the general trends and patterns are comparable and are in good agreement with the nearby Rasova section (Zeeden et al., in press) (Fig. S7.7). Records of χfd from both sections indicate that the period between 50,000 and 40,000 years ago was characterised by enhanced moisture. However, shortly after the Campanian Ignimbrite/Y-5 tephra deposition, χfd records suggest a trend of decreasing humidity over late MIS 3 in the Lower Danube Basin, also observed in fine grain-size fractions (Figs. 7.2 and 7.3). Trends of decreasing humidity and consequently weaker weathering are additionally supported by the decreasing

110

trends of CIA and CPA during the late MIS 3 (Fig. 7.4). The mean annual lake surface temperature of the Black Sea (Wegwerth et al., 2015) does not show any similar trend of decreased humidity and/or temperatures during late MIS 3 (Fig. 7.3). This suggests that a change in trend towards climatic deterioration and progressive continentalization of the Lower Danube Basin is related to a change in atmospheric circulation, rather than in changes in the Black Sea mean annual lake surface temperature. However, the long-term glacial vegetation dynamics inferred from a pollen record from the Black Sea sediment core specify the period from ~40,000-32,000 years BP as a major arid phase (Shumilovskikh et al., 2014). Thus, the general continentalization trend during late MIS 3 is clearly observed in the terrestrial environments surrounding the Black Sea.

During MIS 2, the Lower Danube Basin experienced dry and cold environmental and climatic conditions as indicated by a low fine fractions contribution, decreased χfd values (Fig. 7.2) and low weathering as indicated by CIA and CPA (Fig. 7.4). However, late MIS 2 shows a sharp increase in χfd reaching the highest values in the Holocene, indicating a noticeable increase in regional humidity with the onset of the deglaciation (Fig. 7.2). Accordingly, the Lower Danube Basin experienced relatively mild conditions during middle MIS 3 (50,000- 40,000 years ago), while late MIS 3 was exposed to the general deterioration of the environmental and climatic conditions, with pulses of increased humidity possibly related to interstadial phases. Early MIS 2 and the LGM were generally dry and cold with increased wind intensity, while the deglaciation period is characterised by a sharp increase in humidity and a weakening of wind strength.

Our data indicate that the onset of a general continentalization (less precipitation and colder winters) of climate in the Lower Danube Basin became a persistent feature from ~40,000- 27,000 years (Figs. 7.2-7.4). Nevertheless, the observed continentalization is not only a regional feature of the Lower Danube Basin. The ELSA (Eifel Laminated Sediment Archive) Vegetation-Stack (Sirocko et al., 2016) from Western Europe also indicates a trend of continentalization, culminating as a shift from boreal forest to steppe conditions at ~36,500 years ago (Fig. S7.11). This implies that large areas of Eastern and Western Europe experienced a major environmental change towards open steppe within late MIS 3. The continentalization of major parts of Europe (from east to west) can be explained by the increase in the Fennoscandian ice sheets or with the general increase of the Eurasian high- pressure system. Recent studies report the Fennoscandian ice sheet to a very limited extent 111

before ~35,000 years ago, while the major expansion started only after ~30,000 years ago (Helmens, 2014; Hughes et al., 2016). Moreover, some studies even argue for an ice-free MIS 3 in Fennoscandia (Sarala et al., 2016). Accordingly, an increase in the extent of the Fennoscandian ice sheet was not a forcing mechanism for the observed changes in atmospheric circulation.

It is widely accepted that increases in grain-size seen in records from the Chinese Loess Plateau are linked to a strengthening of the East Asian winter monsoon due to an intensification of the Eurasian high pressure system (the Siberian High) (Ding et al., 1995; Hao et al., 2012). Therefore, the grain-size record from the Chinese loess sequences is a reliable indicator of the Siberian High intensity. The Chinese loess records show a clear trend of increase in grain-size during late MIS 3 (Hao et al., 2012, 2015; Yang and Ding, 2014) (Fig. 7.2), indicating a strong increase in the Siberian High intensity during this time period. The similarities in the grain-size trends between the Lower Danube loess sections discussed in this study and data from the Chinese Loess Plateau over late MIS 3 (Fig. 7.2) support that a common Eurasian atmospheric forcing pattern was responsible for the climatic evolution of these two regions during that time period. Therefore, we argue for an increased influence of the Eurasian high-pressure system (the Siberian High) as the determining factor for palaeoclimate over major parts of Eurasia, including the Lower Danube Basin, during late MIS 3. This for the first time suggests that the Siberian High had a crucial influence on European climate regimes during the time period when the Fennoscandian ice sheets still had a limited extent (Helmens, 2014; Hughes et al., 2016).

The increasing influence of the Siberian High on Europe must have had a strong influence on prevailing air masses from the Atlantic (Westerlies) during the limited extent of the Fennoscandian ice sheet. The speleothem record from the Dim Cave (Ünal-İmer et al., 2015) (Fig. 7.2) suggests a shift of the Westerlies from the European track (NW-SE European trajectory across the Balkans) towards the Mediterranean track (W-E trajectory where the air mass passes over the Mediterranean Sea) during late MIS 3 (Fig. 7.2). This shift in the course of the Westerlies can be explained with an increased influence of the Siberian High on Europe. Increasing air pressure over Eastern and Western Europe during late MIS 3 caused a shift of the Westerlies with lower air pressure to the south.

112

Contrary to observations of general continentalization from the Lower Danube Basin, loess records in the Middle Danube Basin indicate an opposite trend (Fig. 7.3) (Antoine et al., 2009; Basarin et al., 2014; Bokhorst et al., 2011; Hatté et al., 2013; Stevens et al., 2011). The Middle Danube Basin is characterised by an increase in finer particles and in χfd after ~40,000 years ago (Antoine et al., 2009; Basarin et al., 2014; Bokhorst et al., 2011; Stevens et al., 2011), pointing to warmer and more humid conditions during late MIS 3 (Fig. 7.3). Such an opposing climatic evolution between these areas support the interpretation of a southwards shift of the Westerlies after ~40,000 years ago (Ünal-İmer et al., 2015). The prevailing wind track that was dominant before ~40,000 years had a NW-SE trajectory (Ünal-İmer et al., 2015) bringing the colder air masses from the north over the Middle Danube Basin (Fig. 7.5a). A shift towards the Mediterranean track induced warmer and moist air masses that were present in the Mediterranean (Tzedakis et al., 2002, 2006) and the Balkans (Obreht et al., 2016) to reach the Middle Danube Basin in late MIS 3 (Fig. 7.5b). Warm and moist air masses did not reach the Lower Danube Basin due to the orographic obstacle of the mountain chains throughout the Balkans and particularly the Carpathian Mountains (Figs. 7.1 and 7.5b). Summarizing, we highlight that during the periods of limited extent of the Fennoscandian ice sheets the Siberian High played a crucial role on the evolution of prevailing atmospheric circulations and palaeoenvironmental conditions over Europe.

Such atmospheric circulation over Europe may have had an important role for AMH dispersal into Europe. The first AMH appeared in Europe roughly 45,000 to 40,000 years ago (Mellars, 2006; Nigst et al., 2014), and the reasons for the timing of the AMH dispersal have widely been debated. While some studies highlight the continuous presence of native Neanderthals before their demise at around 40,000 years ago (potentially due to the impact of Heinrich event 4 and the Campi Flegrei super-eruption) as an important obstacle for earlier AMH dispersal (Lowe et al., 2012; Marti et al., 2016), others report AMH cultural layers to be present in Europe before that time period (Benazzi et al., 2015; d’Errico and Banks, 2015; Higham et al., 2011). However, none of these studies advanced a robust explanation for the AMH dispersal and observed replacement of other hominid species. Although it is likely that the AMH reached Europe before ~40,000 years ago, the major dispersal of our ancestors and the final demise of the Neanderthals from Europe occurred shortly after ~40,000 years ago (Banks et al., 2008) (although it is suggested that Neanderthals survived in some parts of

113

southern Iberia (Finlayson et al., 2006) and the Balkans (Higham et al., 2006) until ~32,000 years BP).

Figure 7.5. a) Simplified scheme of proposed general atmospheric circulation patterns over Europe during middle MIS 3. Note that the Westerlies were reaching Central and Southeastern Europe via the NW-SE trajectory; b) Simplified scheme of proposed general atmospheric circulation patterns over Europe during late MIS 3. Increased Siberian High influence on Europe had a major influence on Eastern and Western Europe, while the Westerlies shifted to the S-E trajectory, bringing the warmer air masses from the Mediterranean to the Balkans and the Middle Danube Basin. Blue line represents a schematic intensification of Siberian High, green lines show prevailing air masses and red lines represent paths of AMH dispersal in Europe. The map was generated using ArcGIS 10.2.2 (http://www.esri.com/software/arcgis).

114

Based on the similarities of techno-complexes from lithic assemblages, a commonly proposed corridor of AMH dispersal into Europe is over the Levant and Balkans (Benazzi et al., 2015; Hublin, 2015). However, AMH have been demonstrated as already present in western Siberia (Fu et al., 2014) and on the Don River (Anikovich et al., 2007) in present-day Russia roughly around 45,000 years ago. These finds suggest that the AMH were present in Siberia well before the major dispersal into Europe took place. This highlights the possibility of a northern corridor, over the eastern and northern parts of the Black Sea and the Eastern European Plains. According to techno-complexes, this corridor was likely not the main path of AMH dispersal before ~40,000 years. However, the Campi Flegrei super-eruption and the deposition of Campanian Ignimbrite/Y-5 ash at around 39,930 years ago had a strong impact in this region when compared to the Levant, since thick tephra deposits are reported from the Mediterranean to the Eastern European Plains (Marti et al., 2016). The deposition of several centimeters of volcanic ash as modelled by Marti et al. (2016) or the field evidence of tens of centimeters thick ash beds along the Danube and its side valleys (Fitzsimmons et al., 2013; Veres et al., 2013) must have had a devastating impact on flora and fauna in this region. This may permanently annul or decrease the advantage of primary occupation that Neanderthals had over Southeastern and Eastern European Plains. Based on simulations to define eco- cultural niches associated with Neanderthal and AMH adaptive systems during alternating cold and mild phases of MIS 3, it is suggested that during the following Greenland Interstadial 8 (around 38,000 years ago), AMH expansion resulted in competition with which the Neanderthal adaptive system was unable to cope (Banks et al., 2008). Accordingly, since the Campanian Ignimbrite/Y-5 tephra deposition annulled the advantage of primary occupation, AMH had an advantage in reoccupying the areas affected by the tephra deposition. With an increase in the Siberian High intensity after the Campi Flegrei super- eruption, as reported in this study, intensive westwards dispersal of AMH from western Siberia may have taken place over the Eastern European Plains (Fig. 7.5). Here we suggest that although the first AMH may have used the southern corridor over Balkans to enter Europe, during late MIS 3 (in particular after the Campanian Ignimbrite/Y-5 eruption) the corridor over the Eastern European Plains was likely the most dominant one for the AMH dispersal into Europe. The genetic study of the first known European AMH from Peştera cu Oase cave shows that this population did not contribute substantially to later humans in Europe (Fu et al., 2015, 2016), supporting a later arrival of AMH with different origin via the northern corridor. The permanent intensification of the Siberian High during late MIS 3 115

caused a drier and colder environment (especially cold winters) that became unfavourable for the survival of AMH in western Siberia. With the temporal intensification of the Siberian High adverse environmental and climatic conditions prevailed in Eastern European Plains, but the palaeoenvironment in major parts of Europe became an open and fertile steppe (Fig. S7.11) able to sustain large herds of herbivores and their hunters. Such climatic and environmental changes initiated a significant dispersal of AMH from western Siberia and Eurasian interior towards Western Europe.

7.5. Material and methods

To obtain the grain-size data, subsamples of 0.1–0.3 g fine-earth (<2 mm in size) were pre- treated with 0.70 ml of 30 % hydrogen peroxide (H2O2) at 70 °C for 12 hours. This process was repeated until a bleaching of the sediment occurred, but not longer than three days. To keep particles dispersed, the samples were treated with 1.25 ml, 0.1 M sodiumpyrophosphate

(Na4P2O7 * 10H2O) for 12 h (Obreht et al., 2016, 2015). Particle size characteristics were measured with a LS 13320 Laser Diffraction Particle Size Analyser (Beckman Coulter). To calculate the grain-size distribution the Mie theory was used (Fluid RI: 1.33; Sample RI: 1.55; Imaginary RI: 0.1) (Nottebaum et al., 2015; Özer et al., 2010; Schulte et al., 2016).

Bulk samples for environmental magnetism were dried and packed into plastic boxes, and subsequently compressed and fixed with cotton wool to prevent movement of sediment particles during measurement. For Urluia, the majority of samples (323) were collected as oriented samples, using brass tubes and an orientation holder. This way, samples were placed in the diamagnetic boxes directly in the field. The volumetric magnetic susceptibility was measured at frequencies of 300 and 3000 Hz in a static field of 300 mA/m using a Magnon International VSFM. Data were corrected for drift and for the effect of sampling boxes (weak diamagnetism), and normalised to density. Hence, magnetic susceptibility is given as mass 3 specific susceptibility in m /kg. The frequency dependence was calculated as χfd = (χlf – χhf) /

χlf *100 [%] (Buggle et al., 2014; Heller et al., 1991; Obreht et al., 2016; Zeeden et al., in press).

For geochemical analyses, all bulk sediment samples were sieved down to 63 μm and dried at 105 °C for 12 hours. An 8 g quantity of the sieved material was mixed with 2 g Fluxana Cereox wax, homogenised and pressed to a pellet with a pressure of 19.2 MPa for 120 seconds. The measurements were conducted by means of a pre-calibrated method. 116

Samples were analysed for major and trace element abundances with polarization energy dispersive X-ray fluorescence (EDPXRF) using a SpectroXepos.

For investigating glass shard major oxide chemical composition, the tephra sample >(from the visible tephra layer) was sieved and the residual was mounted in epoxy resin, ground and polished. Measurements were made using single-grain, wavelength-dispersive electron microprobe analysis at the Bayerisches GeoInstitut on a Jeol JXA8200 microprobe employing an accelerating voltage of 15 keV, a 6 nA beam current and defocused beam. Order of measured elements (first to last): Na, Si, K, Ca, Fe, Mg, Al, P, Ti, Mn, Cl, with peak counting times averaging 10s for Na, 30s for Si, Al, K, Ca, Fe and Mg, 40s for Ti and Mn, and 60s for P. Precision is estimated at <1–6% (2σ) and 10–25% (2σ) for major and minor element concentrations, respectively.

Luminescence dating was performed on seven samples from the Urluia section. For equivalent dose (De) determination, fine-grained (4-11μm) polymineral samples were measured in a Risø TL/OSL DA 20 reader at the Cologne Luminescence Lab. The post infrared infrared stimulated luminescence (pIRIR) protocol by Thiel et al. (2011) and the central age model (Galbraith et al., 1999) were used. Prior IR stimulation temperature tests, dose recovery tests, residual and fading measurements were conducted. Radionuclide concentrations were measured in a high-purity germanium gamma-ray spectrometer and converted into dose rates. Additional details on the methodology are presented within the Supplementary Material Chapter 7.

117

8. Synthesis

This dissertation presents a brief overview on loess and the results of palaeoclimate research on loess-palaeosol sequences found in Southeastern Europe. Loess is usually defined as an accumulated wind blown dust, however, a clear definition of loess which also pays attention to post-depositional processes is still missing. Nevertheless, it is clear that loess has an aeolian origin, while the loess-like sediments are related to non-aeolian and redeposited sediments with loess structure. Although processes proceeding the dust accumulation are well understood, trapping of the particles, particle immobilisation and other post-depositional processes are still matter of debate. This is due to poor understanding of loessification, process related to post-depositional loess formation. It is unclear whether the loessification is more related to pedogenesis (making loess a soil) or to cementation (making loess a rock). To date, we are not able to make a clear definition of loess, but it is indicated that the further studies have to be directed to reveal the role of particle trapping in different ecosystems, the processes related to the formation of the loess-like structure, and the role of carbonate cementation, pedogenesis and diagenesis. Answering those questions will not only improve our knowledge of loess as a sediment but also give an advanced interpretation of proxies in loess and improve palaeoclimate reconstruction. Although there are still open questions in loess definition, it is important to address and consider all disadvantages of loess as a palaeoclimate archive when reconstructing the past climatic conditions.

In this dissertation four loess-palaeosol sequences were studied in order to reveal past aeolian dynamics of Southeastern Europe during the Late Pleistocene. The aeolian dynamics and palaeoenvironment reconstruction was based on grain-size, environmental magnetism and geochemistry analyses of loess-palaeosol sequences in Orlovat, Stalać, Urluia and Vlasca. The studied sections are distributed over different parts of Southeastern Europe, where the Orlovat section is located in the Middle Danube Basin, the Stalać section in the Balkans, while the Urluia and Vlasca are in the Lower Danube Basin.

The main goals of this dissertation were 1) to improve our knowledge on the relative influence of past regional winds on the palaeoenvironment, 2) and to reconstruct the spatial evolution and the interaction of large scale atmospheric circulations on Southeastern Europe over the Late Pleistocene. Results from the Orlovat section indicate that local specific winds may have a strong influence on the regional climate and the local sedimentological

118

conditions. It is demonstrated that the Košava wind (regional southeast wind in northern Serbia) had a significant influence on climate and sedimentology of the southern Banat region during MIS 5, while its influence during glacial conditions was less expressed or limited. It is shown that topography, geomorphological features and river dynamics also play an important role in sedimentological dynamics. Moreover, major changes in a source area are recorded in the geochemical fingerprints of the Orlovat section. Changes in the source area occurring during MIS 3 clearly changed the sedimentological dynamics. Studies on the Orlovat section indicate the significance of smaller-scale wind systems on the regional climate, but also highlight the importance of a proper understanding of the interactions between the tectonics, river dynamics and possible changes in the source area for a more complete palaeoclimate interpretation.

The reconstruction of past interactions of large-scale atmospheric systems is based on the principle that Southeastern Europe is at the border between Atlantic, continental and Mediterranean climate zones. Consequently, small changes in the relative influence of associated atmospheric systems during past have had a significant impact on the palaeoclimate of the region. However, it is challenging to distinguish which of these climate systems may have dominated during the past. Investigations at the Stalać section enabled a better understanding of the relative influence of the Mediterranean climate over the last 350,000 years. It is shown that the Mediterranean influence was not stable during the past, and a general trend of the weakening of Mediterranean-like climate has been recorded. Although the studied time interval covers a larger time-frame than the Late Pleistocene, it is demonstrated that the Mediterranean climate was limited during MIS 5 to the southernmost part of Southeastern Europe and that for the first time continental climate prevailed over the Balkans during interstadials. Such a change of prevailing climatic zone resulted in a decrease in precipitation due to a stronger influence of continental climate and a change of precipitation maximum from the cold season (winter) to warmer seasons (spring and autumn). This resulted in less snow cover and a decrease in the extent of glaciers. The absence of large glaciers enabled the penetration of relatively warm and humid air from the south and southwest into the interior of the Balkans, leading to a higher humidity during the last glacial, when compared to previous glaciations. Therefore, the observed shift from the Mediterranean to Continental climate likely resulted in a more humid last glacial over the Balkans during MIS 5.

119

The higher humidity related to warmer air masses from the south was mostly limited to the Balkans during the last glacial. The Middle and Lower Danube Basins experienced colder climate, and the influences related to the Continental and Atlantic climate zones were limited only to the Middle and Lower Danube Basins during the last glacial period. Thus, a comparison of the palaeoclimatic signals from the Middle and Lower Basins might enable a better insight into the spatial evolution and interaction of Continental and Atlantic climates.

Results from the Urluia and Vlasce sections in the Lower Danube Basin indicate a decrease in the frequency dependent magnetic susceptibility and an increase in grain-size distribution. These data suggest a general trend of colder conditions and a decreasing precipitation during late MIS 3. The Middle Danube Basin indicates an opposite trend, where the frequency dependent magnetic susceptibility and finer grain particles increase during late MIS 3. Such an opposing trend points to major changes in the interaction between the Atlantic and Continental climate systems during this period. Similarities in palaeoclimatic proxies from the Lower Danube and Chinese loess indicate a common strong influence of the Siberian High. Since the Fennoscandian ice sheets had a limited extent during late MIS 3, the Siberian High had a crucial effect on European palaeoclimate during this time interval. It is demonstrated that a major part of Eastern Europe, over the Lower Danube Basin to Western Europe, experienced a trend of general continentalization. As a consequence, prevailing winds from the Atlantic shifted to the south bringing the warmer and moister air masses from the south into the Balkans and the Middle Danube Basin.

New perspectives of the research and future focus of studies on loess in Southeastern Europe should pay a special attention to two facts. First, the gradual termination of the Mediterranean influence over time cannot be (at least not fully) explained with the previously proposed hypothesis of uplift of surrounding mountains. Although the uplift of the surrounding mountains might play a certain role, it cannot fully explain the decrease in the influence of the Mediterranean climate over time observed in the Danube Basins and the Balkans. Consequently, this might be a feedback from the Northern Hemisphere ice sheet evolution, uplifting tectonic events on the Eurasian scale and/or intensified continental influence across the interior of Eurasia over time. Investigations have to cover a longer time frame than the focus of this dissertation. However, the general idea of re-evaluation of existing data and need for new studies are integrated into this dissertation. Secondly, it is highlighted that a comparison between the palaeoclimatic conditions from three regions in Southeastern Europe 120

(the Middle and Lower Danube Basins and the Balkans) can be successfully applied for reconstruction of the interaction of large-scale atmospheric systems as big as the European scale. However, this requires the advances in the field of age control, especially for the longer time intervals.

Results from this dissertation pave the way for new ideas, hypotheses and discussions. Observed different past climatic conditions in the Central Balkans and the Danube Basins, and the past sharp atmospheric boundary of those regions can improve our understanding of the Balkan Peninsula as an important Quaternary refugium. Particularly, the enhanced humidity during the last glacial explains such a role of the Balkan Peninsula during this time interval. Moreover, this dissertation fills some gaps in our knowledge concerning the dispersal of Anatomically Modern Humans to Europe. The European palaeoclimatic conditions and atmospheric circulations had substantially changed during the late MIS 3 (the period when the Anatomically Modern Humans appeared in Europe), in particular because of the increased influence of the Siberian High. It is proposed that with the permanent intensification of the Siberian High during the late MIS 3, the Eurasian interior transformed into a dry and cold environment that became unfavourable for the survival of the Anatomically Modern Humans. On the contrary, the palaeoenvironment in major parts of Europe became an open and fertile steppe able to sustain large herds of herbivores. Such conditions might have triggered massive migrations of Anatomically Modern Humans to Europe via a northern corridor.

121

9. Abstract

Loess is a valuable archive of past climatic and environmental conditions. Southeastern European thick loess-palaeosol sequences have preserved a quasi-continuous record of glacial-interglacial cycles, reaching as far as the Early Pleistocene. The Middle and Lower Danube Basins are lowland areas with thick loess plateaus comprising several glacial- interglacial cycles. Those areas have been the focus of recent research on the Quaternary in Europe, which has resulted in a better understanding of the Pleistocene palaeoenvironmental and palaeoclimatic signals. However, while some local and regional palaeoclimatic information is available, large-scale forcing mechanisms and the climatic conditions responsible for loess formation have not been revealed. Contrary to the Middle and Lower Danube Basins, the interior of the Balkans is poorly investigated. Lack of information of past climatic and environmental conditions over the Balkans represents an additional barrier to the understanding of large-scale atmospheric circulations over Southeastern Europe.

The aim of this dissertation is to reconstruct aeolian dynamics and climatic conditions over Southeastern Europe during the Late Pleistocene. Before the reconstruction of the past climatic conditions, a definition of loess and its evolution through time is given. Loess is usually defined as a wind-blown dust that after deposition undergoes a special process called loessification. The lack in knowledge and understanding of postdepositional processes during loessification make a clear definition of loess difficult. For now, it is proposed to define loess as a mainly aeolian dust deposit that has experienced loessification, resulting in loess-like aggregation.

The sections studied for palaeoclimatic reconstruction are located in different regions in Southeastern Europe. The Orlovat section is located in the Middle Danube Basin (Vojvodina, Serbia), the Stalać section is in the Central Balkans (Central Serbia), and the Urluia and Vlasca sections are situated in the Lower Danube region (Dobrogea, Romania). Regional palaeoclimate conditions of these three areas have been reconstructed. It is shown that past climatic and environmental conditions have been very diverse among different regions of Southeastern Europe. The climate evolution was dynamic and different even within smaller regions, such as parts of the southern Middle Danube Basin. The southeast Košava wind, a regional wind of the southern Banat region, had a strong impact on the environmental conditions over the Banat region, particularly during MIS 5 and the Holocene. The

122

sedimentological dynamics were determined by the relative influence of local winds and particularly changes in the source areas.

Palaeoclimatic evolution over the Central Balkans was a very dynamic process over the past 350,000 years. Mediterranean-like climate was very prominent during the Middle Pleistocene. However, the influence of Mediterranean climate weakened over time, and was finally replaced by continental climate influence during MIS 5. The reported shift in a dominant atmospheric mode over this area during MIS 5, and associated changes in precipitation, resulted in warmer and more humid conditions over the following glacial cycle when compared to previous glaciations. Such a palaeoclimatic evolution during the Late Pleistocene was characteristic only for the Balkans area, and the climatic conditions in other part of Southeastern Europe were colder and drier.

During the Late Pleistocene the Mediterranean-like climate was not present in Southeastern Europe, although the relative influence of the warmer air masses from the south was present in the Central Balkans. Accordingly, climatic influences over the Middle and Lower Danube Basins were associated with the Atlantic and continental climates only. However, the influence of those climates did not equally affect those regions over time. Especially between 40,000 and 27,000 years ago, the Lower Danube Basin was under a stronger influence of continental climate conditions (with especially colder winters). On the contrary, the Middle Danube Basin was under the influence of Atlantic climate and milder climatic conditions. Such opposing trends indicate major large-scale atmospheric changes in Europe during the mentioned time interval. An explanation for the opposing trends and general changes in the European large-scale atmospheric systems is given. During this time interval (late MIS 3 - 40,000 to 27,000 years ago) the intensification of the Siberian High had a crucial influence on European climate causing a general continentalization of major parts of Europe. An area from the Lower Danube Basin to Western Europe experienced a trend of increasing air pressure because of an intensification of the Siberian High, causing the southwards shift of the Westerlies. The Westerlies from the south had been bringing warmer and more humid air masses into the Balkans and the Middle Danube Basin. This highlights the fact that during the limited extent of the Fennoscandian ice sheets the Siberian High played a crucial role in the evolution of prevailing atmospheric circulations and palaeoenvironmental conditions in Europe.

123

Based on presented large-scale atmospheric systems evolution in Europe, this dissertation also provides the climatic and environmental context for the Anatomically Modern Human dispersal into Europe. It is highlighted that with the intensification of the Siberian High during late MIS 3 adverse environmental and climatic conditions prevailed in Eastern European Plains, but the palaeoenvironment in major parts of Europe became an open and fertile steppe able to sustain large herds of herbivores and their hunters. Thus, a possible intensive westwards dispersal of Anatomically Modern Human from western Siberia may have taken place over the Eastern European Plains, especially after the Campanian Ignimbrite/Y-5 tephra deposition.

124

10. Zusammenfassung

Löss ist ein wertvolles Archiv vergangener Klima- und Umweltbedingungen. Die mächtigen südosteuropäischen Löss-Paläobodensequenzen, insbesondere in den mittleren und unteren Donaubecken, sind quasi-kontinuierliche Archive der Glazial-Interglazial Zyklen, die bis in das frühe Pleistozän zurück reichen. Rezente Bestrebungen der Quartärforschung in Europa konzentrieren sich auf diese Gebiete, um zu einem besseren Verständnis der pleistozänen Paläoumwelt und der Paläoklimasignale beizutragen. Bisher konnten jedoch nur einige lokale und regionale Paläoklimainformationen gewonnen werden, während die großräumige Klimafaktoren, die zur Lössbildung führen nicht aufgedeckt werden konnten. Im Gegensatz zu den mittleren und unteren Donaubecken sind die zentralen Regionen des Balkans bislang kaum untersucht worden. Aufgrund dieser Datenlücken sind auch die Veränderungen in der atmosphärischen Zirkulation in Südosteuropa bisher nur schlecht verstanden.

Das Ziel dieser Doktorarbeit ist die Rekonstruktion äolischer Dynamiken und klimatischer Bedingungen in Südosteuropa während des Spätpleistozäns. Vor der Rekonstruktion vergangener klimatischer Bedingungen wird eine Definition von Löss und derer Entwicklung im Laufe der Zeit erörtert. Löss wird gewöhnlich als ein von Wind transportiertes Sediment definiert, welches nach Akkumulation post-sedimentäre Überprägung, die sogenannte Lössifikation, erfährt. Der Mangel an Kenntnissen und Verständnis der während der Lössifikation ablaufenden post-sedimentären Prozesse macht eine klare Definition von Löss schwierig. In dieser Arbeit wird Löss als hauptsächlich äolisch abgelagerte Partikel definiert, die nach der Akkumulation lössifiziert wurden und aus denen sich eine lössähnliche Aggregation bildet.

Die für die paläoklimatische Rekonstruktion untersuchten Löss-Paläobodensequenzen befinden sich in unterschiedlichen Regionen Südosteuropas. Das Profil Orlovat befindet sich im mittleren Donaubecken (Vojvodina, Serbien), die Stalać Sequenz ist im zentralen Balkan (Zentralserbien) und die Sequenzen Urluia und Vlasca befinden sich in der unteren Donauregion (Dobrogea, Romänien). Die Lösssequenzen dienen als Proxidaten zur Rekonstruktion der Paläoklima- und Paläoumweltbedingungen. Es hat sich gezeigt, dass die vergangenen Klima- und Umweltbedingungen in den unterschiedlichen Regionen in Südosteuropa sehr verschieden waren. Die klimatische Entwicklung war hoch variabel und verschieden selbst in kleineren Gebieten, wie dem südlichen mittleren Donaubacken. Der

125

Košava Wind, ein regionales Windsystem des südlichen Banats aus südöstlicher Richtung, hatte einen großen Einfluss auf die Umweltbedingungen des Banats, besonders während des MIS 5 und des Holozäns. Die sedimentologischen Dynamiken waren von dem relativen Einfluss der regionalen Winde und besonders von Änderungen der Liefergebiete bestimmt.

Die paläoklimatischen Entwicklungen über dem zentralen Balkan waren während der letzten 350,000 Jahre sehr dynamisch. Während des Mittelpleistozäns herrschte vor allem mediterranähnliches Klima vor. Im Laufe der Zeit nahm der mediterrane Einfluss ab und während des MIS 5 herrschten letztendlich kontinentalere Klimabedingungen vor. Der aufgezeichnete Wechsel der dominierenden atmosphärischen Bedingungen über der Region während des MIS 5 und dadurch verursachte Änderungen der Niederschlagsmengen resultierten in wärmeren und humideren Bedingungen während des darauffolgenden Glazials im Vergleich zu den vorangegangenen. Eine solche paläoklimatische Entwicklung während des Spätpleistozäns war ausschließlich für die Balkanregion charakteristisch während die Klimabedingungen in anderen Teilen Südosteuropas kälter und trockener waren.

Trotz des Einflusses warmer Luftmassen von Süden auf den zentralen Balkan, gab es während des Jungpleistozäns in Südosteuropa kein mediterranes oder mediterranähnliches Klima mehr. Stattdessen konnte der klimatische Einfluss im mittleren und unteren Donaubecken mit atlantischem und kontinentalem Klima in Verbindung gebracht werden. Nichtsdestotrotz wurden diese Regionen nicht gleichmäßig und gleichzeitig von diesen Klimazonen beeinflusst. Das untere Donaubecken stand vor allem zwischen 40.000 und 27.000 Jahren vor heute unter einem stärkeren kontinentalen Einfluss (mit besonders kalten Wintern). Im Gegensatz dazu herrschten im mittleren Donaubecken währenddessen mildere klimatische Bedingungen unter atlantischem Klima vor. Diese gegensätzlichen Tendenzen deuten auf bedeutende und großräumige atmosphärische Veränderungen in Europa während des angegebenen Zeitintervalls hin. Einen entscheidenden und stärkeren Einfluss auf das europäische Klima während dieser Zeit hatte ein ausgedehnteres sibirisches Hochdruckgebiet. Diese Intensivierung führte zur Kontinentalisierung großer Teile Europas. Durch eine Zunahme des Luftdruckes zwischen dem unteren Donaubecken und Westeuropa wurden die Westwinde nach Süden verschoben, was den Transport wärmerer und feuchterer Luftmassen in den Balkan und das mittlere Donaubecken begünstigte. Dies hebt die Rolle des sibirischen Hochs auf die Entwicklung der vorherrschenden atmosphärischen Zirkulation und

126

Paläobedingungen in Europa hervor, insbesondere zu den Zeiträumen, in denen das fennoskandinavische Inlandeis nur geringe Ausmaße besaß.

Basierend auf der präsentierten großräumigen atmosphärischen Systementwicklung in Europa liefert die vorliegende Dissertation auch Hinweise auf den klimatischen und den ökologischen Kontext für die Ausbreitung des anatomisch modernen Menschen nach Europa. Es wird betont, dass die Intensivierung des Sibirischen Hochs während des späten MIS 3 zwar ungünstige Umwelt- und Klimabedingungen in den osteuropäischen Ebenen verursachte, in den meisten andere Regionen Europas jedoch die Paläoumweltbedingungen durch offene und fruchtbare Steppen und Lößtundren gekennzeichnet waren. Diese bildeten die Lebensgrundlage für große Herden von Pflanzenfressern und damit auch für eiszeitlicher Jäger- und Sammlerkulturen.

127

11. Acknowledgments and author contributions

The past three years were a wonderful experience for me, both on a scientific and a personal level. I shared my time with great people that had a remarkable influence on my scientific and personal development. Here I would like to acknowledge them.

I would like to express my deepest appreciation and respect to my supervisor Prof Dr Frank Lehmkuhl. Thanks to his generous and continuous support during the past three years, it was a real pleasure to extend my knowledge. I am very grateful that I had an opportunity to work with him, and I am thankful for his guidance and his directions in my scientific development.

My special thanks go to Dr Christian Zeeden, for being a tutor, field companion and a big support, but most of all a great friend. Intensive discussion, great suggestions and remarks significantly (I know you will say not to use this word if it cannot be calculated) changed my way of thinking. I am sure we will have many more days to discuss about science and life with hopefully some Tokaj wine.

I thank Janina Bösken for her wonderful friendship and for being such a great colleague. It was a great pleasure to work with her on the field, to discuss after and to find solutions for many problems our data made to us. I thank Lydia Krauß for being the best office partner one can imagine. Having her as a friend and a colleague made my working days better and full of laughter (and cookies). To both of them I own my deepest gratefulness for not letting me sink into the floods of uninteresting obligations and administration.

I am very grateful to Dr Ulrich Hambach, from whom I have learned the most during the past three years. I would like to thank him for his time for our extensive discussion over a phone, Skype, emails and in person. I thank Dr Daniel Veres, who was a great inspiration on a field. I had an unforgettable time together with these guys.

My appreciation goes to Dr Tom Stevens for his time and advices that sincerely helped in my scientific development.

Thanks to all my former and present colleagues from the RWTH Aachen University, especially to the PGG group. Among them, I would like especially to thank Veit Nottebaum (also for making me a Frankfurt fan), Kaifeng Yu, Philipp Schulte, Jörg Zens, David Loibl, Eileen Eckmeier and Georg Stauch for great and constructive discussions. My sincere thanks

128

also go to Marianne Dohms for her patience with me in the lab and for the processing of thousands of samples. I am grateful to all student helpers (Alex, Arndt, Arne, Berit, Dimi, Judith, Sophia, Steffi, Benoit among others), without them it would be almost impossible to achieve everything we did. I am especially grateful to Kathrin Emndson, Beate Wieland and Stephan Pötter (my field roommate and accomplice in watching the trashy shark movies) for the unforgettable time on the field. I also thank Janek Walk for his help with some figures and for our competition in darts. Thanks to Anja Knops for her patience and her help with administrative things.

I thank Christa Loibl, the first person I met in Aachen, a great partner on the field and most of all, a wonderful friend.

I would also like to thank to all members of CRC 806 project. I thank to DFG and CRC 806 project for making possible to do this dissertation. Special thanks go to Wei Chu, who gave me a chance to feel like an archaeologist for a certain time. I wish to give my acknowledgement to the IRTG and all its members for a great time.

Thanks go to Nikola Bačević, Zoran Perić and Nemanja Tomić for their great company and help on the field.

I sincerely thank Dragana for her love, support, understanding, and most of all, for being my inspiration during hard times. She was constantly and unselfishly supporting me and with her by my side all problems were easier to overcome.

I thank to my brothers Rastko and Andrej for their support, love and happiness they brought into my life.

I own my deepest and sincere gratefulness to my mother Zorica and to Slobodan, who loved me, cared for me and were there every time I needed them. Their love guided me through life and made me who I am. Thank you!

Manuscripts acknowledgments and author contributions:

Chapter 4: The authors thank X. Suarez-Villagran for several stylistic corrections on the manuscript. O.I. acknowledges the funding of CRC 806 “Our way to Europe”, subproject B1 “The Eastern Trajectory”, supported by the DFG (Deutsche Forschungsgemeinschaft). We are grateful to I. Smalley, who inspired this paper and to A. Makeev for very useful reviewer comments. 129

Author contributions: T.S. and I.O. designed the study and wrote the manuscript.

Chapter 5: The investigations were carried out in the frame of the CRC 806 “Our way to Europe”, subproject B1 The Eastern Trajectory: “Last Glacial Paleogeography and Archeology of the Eastern Mediterranean and of the Balkan Peninsula”, supported by the DFG (Deutsche Forschungsgemeinschaft). We are grateful to Nemanja Tomić, Dragan Popov and Rastko Marković for the help in the field. We thanks to Marianne Dohms for her help with laboratory analysis and Janina Bösken for commenting on the manuscript.

Author contributions: I.O. and F.L. designed the study. I.O. and P.S. supervised and performed laboratory analyses. I.O. wrote the main text of the manuscript. All authors significantly contributed to the interpretation of the data and provided significant input to the manuscript. All authors revised the manuscript.

Chapter 6: We thank the editor Junsheng Nie and three unknown reviewers for thorough and constructive reviews that significantly improved this paper. The investigations were carried out in the frame of the CRC 806 “Our way to Europe”, subproject B1 The Eastern Trajectory: “Last Glacial Paleogeography and Archeology of the Eastern Mediterranean and of the Balkan Peninsula”, supported by the DFG (Deutsche Forschungsgemeinschaft, grant number INST 216/596-2). We are grateful to Christa Loibl, Nemanja Tomić and Rastko Marković for help in the field, Marianne Dohms for her help with laboratory analyses and Janek Walk for his help with Figure 6.1. We would also like to express our gratitude to the staff of the Bayerisches GeoInstitut (BGI, Bayreuth, Germany) for assistance with the sample preparations and microprobe analyses. We thank Randy Schaetzl and Jussi Meriluoto for commenting on the manuscript.

Author contributions: I.O., S.B.M. and F.L. designed the study. I.O., S.B.M., F.L., U.H., C.Z., N.B. and M.B.G. undertook the sampling of profiles. I.O., C.Z., F.L. and U.H. supervised and performed laboratory analyses. I.O. wrote the main text of the manuscript. All authors significantly contributed to the interpretation of the data and provided significant input to the manuscript. All authors revised the manuscript.

Chapter 7: The investigations were carried out in the frame of the CRC 806 “Our way to Europe”, subproject B1, The Eastern Trajectory: “Last Glacial Paleogeography and Archeology of the Eastern Mediterranean and of the Balkan Peninsula”, supported by the

130

DFG (Deutsche Forschungsgemeinschaft, grant number INST 216/596-2). D.V. acknowledges support from project PN-II-ID-PCE- 2012-4-0530 ‘Millennial-scale geochemical records of anthropogenic impact and natural climate change in the Romanian Carpathians’. We thank to Alida Timar-Gabor, Dušan Borić and Zorica Svirčev for their useful comments on the manuscript. We are grateful to Stephan Pötter, Beate Wieland, Kathrin Emunds and Philipp Jäger for help in the field, and Marianne Dohms for her help with laboratory analyses. We would also like to express our gratitude to the staff of the Bayerisches GeoInstitut (BGI, Bayreuth, Germany) for assistance with the sample preparations and microprobe analyses.

Author contributions: U.H., D.V., I.O. and F.L. designed the study. D.V., I.O., U.H., C.Z., J.B., S.M. and F.L. undertook the sampling of profiles. I.O., C.Z. and U.H. supervised and performed gain-size and environmental magnetic analyses. D.V. and U.H. performed microprobe analyses and interpreted the data. J.B., N.K., D.B., C.B. did luminescence dating and interpretation of the data. I.O. wrote the main text of the manuscript. All authors significantly contributed to the interpretation of the data and provided significant input to the manuscript. All authors revised the manuscript.

131

12. References

Allen, J.R.L., Thornley, D.M., 2004. Laser granulometry of Holocene estuarine silts: effects of hydrogen peroxide treatment. The Holocene 14, 290–295.

An, Z., Kukla, G.J., Porter, S.C., Xiao, J., 1991. Magnetic susceptibility evidence of monsoon variation on the Loess Plateau of central China during the last 130,000 years. Quat. Res. 36, 29–36.

Andersen, K.K., Azuma, N., Barnola, J.-M., Bigler, M., Biscaye, P., Caillon, N., Chappellaz, J., Clausen, H.B., Dahl-Jensen, D., Fischer, H., Flückiger, J., Fritzsche, D., Fujii, Y., Goto-Azuma, K., Grønvold, K., Gundestrup, N.S., Hansson, M., Huber, C., Hvidberg, C.S., Johnsen, S.J., Jonsell, U., Jouzel, J., Kipfstuhl, S., Landais, A., Leuenberger, M., Lorrain, R., Masson-Delmotte, V., Miller, H., Motoyama, H., Narita, H., Popp, T., Rasmussen, S.O., Raynaud, D., Rothlisberger, R., Ruth, U., Samyn, D., Schwander, J., Shoji, H., Siggard-Andersen, M.-L., Steffensen, J.P., Stocker, T., Sveinbjörnsdóttir, A.E., Svensson, A., Takata, M., Tison, J.-L., Thorsteinsson, T., Watanabe, O., Wilhelms, F., White, J.W.C., 2004. High-resolution record of Northern Hemisphere climate extending into the last interglacial period. Nature 431, 147–151.

Anechitei-Deacu, V., Timar-Gabor, A., Fitzsimmons, K.E., Veres, D., Hambach, U., 2014. Multi-method luminescence investigations on quartz grains of different sizes extracted from a loess section in Southeast Romania interbedding the Campanian Ignimbrite ash layer. Geochronometria 41, 1–14.

Anikovich, M.V., Sinitsyn, A.A., Hoffecker, J.F., Holliday, V.T., Popov, V.V., Lisitsyn, S.N., Forman, S.L., Levkovskaya, G.M., Pospelova, G.A., Kuz’mina, I.E., Burova, N.D., Goldberg, P., Macphail, R.I., Giaccio, B., Praslov, N.D., 2007. Early Upper Paleolithic in Eastern Europe and Implications for the Dispersal of Modern Humans. Science 315, 223–226.

Ankjærgaard, C., 2010. Understanding optically stimulated charge movement in quartz and feldspar using time-resolved measurements. Ph.D. thesis, Technical University of Denmark. Risø National Laboratory for Sustainable Energy.

132

Antoine, P., Rousseau, D.D., Zöller, L., Lang, A., Munaut, A.V., Hatté, C., Fontugne, M., 2001. High-resolution record of the last Interglacial-glacial cycle in the Nussloch loess-palaeosol sequences, Upper Rhine Area, Germany. Quat. Int. 76-77, 211-229.

Antoine, P., Rousseau, D.D., Hatté, C., Zöller, L., Lang, A., Fontugne, M., Moine, O., 2003. Événements éoliens rapides en context loessique: l’exemple de la séquence du Pléniglaciaire supérieur weichselien de Nussloch (Vallée du Rhin-Allemagne). Quaternaire 13, 199–208.

Antoine, P., Rousseau, D.-D., Fuchs, M., Hatté, C., Gauthier, C., Marković, S.B., Jovanović, M., Gaudenyi, T., Moine, O., Rossignol, J., 2009. High-resolution record of the last climatic cycle in the southern Carpathian Basin (Surduk, Vojvodina, Serbia). Quat. Int. 198, 19–36.

Antoine, P., Rousseau, D.D., Moine, O., Kunesch, S., Hatté, C., Lang, A., Tissoux, H., Zöller, L., 2009. Rapid and cyclic aeolian deposition during the Last Glacial in European loess: a high-resolution record from Nussloch, Germany. Quat. Sci. Rev. 28, 2955- 2973.

Antoine, P., Rousseau, D.D., Degeai, J.P., Moine, O., Lagroix, F., Kreutzer, S., Fuchs, M., Hatté, C., Gauthier, C., Svoboda, J., Lisá, L., 2013. High-resolution record of the environmental response to climatic variations during the Last Interglacial-Glacial cycle in Central Europe: the loess-palaeosol sequence of Dolní Vestonice (Czech Republic). Quat. Sci. Revi. 67, 17-38.

Assallay, A.M., Rogers, C.D.F., Smalley, I.J., 1997. Formation and collapse of metastable particle packings and open structures in loess deposits. Engineering Geology 48, 101- 115.

Assallay, A.M., Rogers, C.D.F., Smalley, I.J., Jefferson, I.F., 1998. Silt: 2-62 mu m, 9-4 phi. Earth Sci. Revi. 45, 61-88.

Badura, J., Jary, Z., Smalley, I., 2013. Sources of loess material for deposits in Poland and parts of Central Europe: The lost Big River. Quat. Int., Closing the gap - North Carpathian loess traverse in the Eurasian loess belt 6th Loess Seminar, Wrocław, Poland Dedicated to Prof. Henryk Maruszczak 296, 15–22.

133

Băltean, I., 2011. The palaeolithic in Banat. In: Drașovean, F., Jovanović, B. (Eds.), The Prehistory of Banat, the Palaeolithic and Mesolithic, București, vol. 1, pp. 19–76.

Barbu, S., Moisa, C., Mircov, V.D., 2009. Kosava – a Baneten wind. Res. J. Agric. Sci. 41, 361–365.

Banks, W.E., d’Errico, F., Peterson, A.T., Kageyama, M., Sima, A., Sánchez-Goñi, M.-F., 2008. Neanderthal Extinction by Competitive Exclusion. PLOS ONE 3, e3972.

Basarin, B., Vandenberghe, D.A.G., Marković, S.B., Catto, N., Hambach, U., Vasiliniuc, S., Derese, C., Rončević, S., Vasiljević, D.A., Rajić, L., 2011. The Belotinac section (Southern Serbia) at the southern limit of the European loess belt: Initial results. Quat. Int., The Second Loessfest (2009) 240, 128–138.

Basarin, B., Buggle, B., Hambach, U., Marković, S.B., Dhand, K.O., Kovačević, A., Stevens, T., Guo, Z., Lukić, T., 2014. Time-scale and astronomical forcing of Serbian loess– paleosol sequences. Glob. Planet. Change 122, 89–106.

Bell, W.T., 1980. Alpha dose attenuation in quartz grains for thermoluminescence dating. Anc. TL 12, 4–8.

Benazzi, S., Douka, K., Fornai, C., Bauer, C.C., Kullmer, O., Svoboda, J., Pap, I., Mallegni, F., Bayle, P., Coquerelle, M., Condemi, S., Ronchitelli, A., Harvati, K., Weber, G.W., 2011. Early dispersal of modern humans in Europe and implications for Neanderthal behaviour. Nature 479, 525–528.

Benazzi, S., Slon, V., Talamo, S., Negrino, F., Peresani, M., Bailey, S.E., Sawyer, S., Panetta, D., Vicino, G., Starnini, E., Mannino, M.A., Salvadori, P.A., Meyer, M., Pääbo, S., Hublin, J.-J., 2015. The makers of the Protoaurignacian and implications for Neandertal extinction. Science 348, 793–796.

Berg, L.S., 1916. The origin of loess (in Russian). Izvestiya Russkogo Geograficheskogo Obshchestva 52, 579-646.

Berg, L.S., 1964. Loess as a Product of Weathering and Soil Formation. Israel Program for Scientific Translations, Jerusalem.

134

Blott, S.J., Pye, K., 2012. Particle size scales and classification of sediment types based on particle size distributions: Review and recommended procedures. Sedimentology 59, 2071–2096.

Boguckyj, A.B., Łanczont, M., Lacka, B., Madeyska, T., Sytnyk, O., 2009. Age and the palaeoenvironment of the West Ukrainian palaeolithic: the case of Velykyi Glybochok multi-cultural site. J. Archaeol. Sci. 36, 1376-1389.

Bokhorst, M.P., Beets, C.J., Marković, S.B., Gerasimenko, N.P., Matviishina, Z.N., Frechen, M., 2009. Pedo-chemical climate proxies in Late Pleistocene Serbian–Ukranian loess sequences. Quat. Int., Loess in the Danube Region and Surrounding Loess Provinces: The Marsigli Memorial Volume 198, 113–123.

Bokhorst, M.P., Vandenberghe, J., Sümegi, P., Łanczont, M., Gerasimenko, N.P., Matviishina, Z.N., Marković, S.B., Frechen, M., 2011. Atmospheric circulation patterns in central and eastern Europe during the Weichselian Pleniglacial inferred from loess grain-size records. Quat. Int., Loess in Eurasia 234, 62–74.

Bösken, J., Klasen, N., Zeeden, C., Obreht, I., Marković, S.B., Hambach, U., Lehmkuhl, F., 2017. New luminescence-based geochronology framing the last two glacial cycles at the southern limit of European Pleistocene loess in Stalać (Serbia). Geochronometria 44, 150-161.

Boulton, G.S., 1978. Boulder shapes and grain-size distributions of debris as indicators of transport paths through a glacier and till genesis. Sedimentology 25, 773-799.

Brennan, B.J., 2003. Beta doses to spherical grains. Radiat. Meas., Proceedings of the 10th international Conference on Luminescence and Electron-Spin Resonance Dating (LED 2002) 37, 299–303.

Bronger, A., 2003. Correlation of loess-paleosol sequences in East and Central Asia with SE Central Europe: towards a continental Quaternary pedostratigraphy and paleoclimatic history. Quat. Int. 106, 11-31.

Buggle, B., Glaser, B., Hambach, U., Gerasimenko, N., Marković, S., 2011. An evaluation of geochemical weathering indices in loess–paleosol studies. Quat. Int. 240, 12–21.

135

Buggle, B., Glaser, B., Zöller, L., Hambach, U., Marković, S., Glaser, I., Gerasimenko, N., 2008. Geochemical characterization and origin of Southeastern and Eastern European loesses (Serbia, Romania, Ukraine). Quat. Sci. Rev. 27, 1058–1075.

Buggle, B., Hambach, U., Glaser, B., Gerasimenko, N., Marković, S., Glaser, I., Zöller, L., 2009. Stratigraphy, and spatial and temporal paleoclimatic trends in Southeastern/Eastern European loess–paleosol sequences. Quat. Int., Lower Latitudes Loess-Dust Transport Past and Present 196, 86–106.

Buggle, B., Hambach, U., Kehl, M., Marković, S.B., Zöller, L., Glaser, B., 2013. The progressive evolution of a continental climate in southeast-central European lowlands during the Middle Pleistocene recorded in loess paleosol sequences. Geology 41, 771– 774.

Buggle, B., Hambach, U., Müller, K., Zöller, L., Marković, S.B., Glaser, B., 2014. Iron mineralogical proxies and Quaternary climate change in SE-European loess–paleosol sequences. CATENA117, 4–22.

Bukurov, B., 1982. Sintetička razmatranja geomorfoloških problema na teritoriji Vojvodine (synthetic investigation of geomorphological problems in Vojvodina). VANU, Novi Sad (in Serbian).

Burchfiel, B.C., Nakov, R. Dumurdzanov, N., Papanikolaou, D., Tzankov, T. Serafimovski, T., King, R.W., Kotzev, V., Todosov, A., Nurce, B. 2008. Evolution and dynamics of the Cenozoic tectonics of the South Balkan extensional system. Geosphere 4, 919-938.

Buurman, P., Pape, T., Muggler, C.C. 1997. Laser grain size determination in soil genetic studies 1. Practical problems. Soil Science 162, 211‐218.

Buylaert, J.-P., Jain, M., Murray, A.S., Thomsen, K.J., Thiel, C., Sohbati, R., 2012. A robust feldspar luminescence dating method for Middle and Late Pleistocene sediments: Feldspar luminescence dating of Middle and Late Pleistocene sediments. Boreas 41, 435–451.

Buylaert, J.-P., Thiel, C., Murray, A., Vandenberghe, D., Yi, S., Lu, H., 2011. IRSL and post- IR IRSL residual doses recorded in modern dust samples from the Chinese Loess Plateau. Geochronometria 38, 432–440.

136

Caseldine, C.J., Turney, C., Long, A.J., 2010. IPCC and palaeoclimate – an evolving story? J. Quat. Sci. 25, 1–4.

Catt, J.A., 2001. The agricultural importance of loess. Earth Sci. Rev. 54, 213-229.

Cilek, V., 2001. The loess deposits of the Bohemian Massif: silt provenance, palaeometeorology and loessification processes. Quat. Int. 76-77, 123-128.

Civetta, L., Orsi, G., Pappalardo, L., Fisher, R.V., Heiken, G., Ort, M., 1997. Geochemical zoning, mingling, eruptive dynamics and depositional processes — the Campanian Ignimbrite, Campi Flegrei caldera, Italy. J. Volcanol. Geotherm. Res. 75, 183–219.

Crouvi, O., Amit, R., Enzel, Y., Gillespie, A.R., 2010. Active sand seas and the formation of desert loess. Quat. Sci. Rev. 29, 2087-2098.

Constantin, D., Timar-Gabor, A., Veres, D., Begy, R., Cosma, C., 2012. SAR-OSL dating of different grain-sized quartz from a sedimentary section in southern Romania interbedding the Campanian Ignimbrite/Y5 ash layer. Quat. Geochronol. 10, 81–86.

Cullers, R.L., 2000. The geochemistry of shales, siltstones and sandstones of Pennsylvanian- Permian age, Colorado, USA: implications for provenance and metamorphic studies. Lithos 51, 181–203. d’Errico, F., Banks, W.E., 2015. Tephra studies and the reconstruction of Middle-to-Upper Paleolithic cultural trajectories. Quat. Sci. Rev., Synchronising Environmental and Archaeological Records using Volcanic Ash Isochrons 118, 182–193. de Boer, B., Lourens, L.J., van de Wal, R.S.W., 2014a. Persistent 400,000-year variability of Antarctic ice volume and the carbon cycle is revealed throughout the Plio-Pleistocene. Nat. Commun. 5, 2999. de Boer, B., Stocchi, P., van de Wal, R.S.W., 2014b. A fully coupled 3-D ice-sheet–sea-level model: algorithm and applications. Geosci Model Dev 7, 2141–2156.

Derbyshire, E., 2001. Geological hazards in loess terrain, with particular reference to the loess regions of China. Earth Sci. Rev. 54, 231-260.

Derbyshire, E., 2003. Loess, and the dust indicators and records of terrestrial and marine palaeoenvironments (DIRTMAP) database. Quat. Sci. Rev. 22, 1813-1819. 137

Derbyshire, E., Mellors, T.W., 1988. Geological and geotechnical characteristics of some loess and Loessic soils from China and Britain - a comparison. Engineering Geology 25, 135-175.

Derbyshire, E., Kemp, R.A., Meng, X., 1997. Climate change, loess and paleosols: proxy measures and resolution in North China. J. Geol. Soc. Lond. 154, 793–805.

De Vivo, B., Rolandi, G., Gans, P.B., Calvert, A., Bohrson, W.A., Spera, F.J., Belkin, H.E., 2001. New constraints on the pyroclastic eruptive history of the Campanian volcanic Plain (Italy). Mineral. Petrol. 73, 47–65.

Dijkstra, T.A., Rogers, C.D.F., Smalley, I.J., Derbyshire, E., Jin, L.Y., Min, M.X., 1994. The loess of North-Central China - geotechnical properties and their relation to slope stability. Engineering Geology 36, 153-171.

Dimitrijević, M.D. (Ed.) 2002: Geological Atlas of Serbia. Serbian Ministry of Natural Resources and Environmental Protection, Beograd.

Ding, Z., Liu, T., Rutter, N.W., Yu, Z., Guo, Z., Zhu, R., 1995. Ice-Volume Forcing of East Asian Winter Monsoon Variations in the Past 800,000 Years. Quat. Res. 44, 149–159. doi:10.1006/qres.1995.1059

Ding, Z.L., Ranov, V., Yang, S.L., Finaev, A., Han, J.M., Wang, G.A., 2002. The loess record in southern Tajikistan and correlation with Chinese loess. Earth Planet. Sci. Lett. 200, 387–400.

Doeglas, D.J., 1968. Grain-size indices, classification and environment. Sedimentology 10, 83–100.

Dokuchaev, V.V., 1883. Russian Chernozem (In Russian). Independent Society for Economics, St. Petersburg.

Durcan, J.A., King, G.E., Duller, G.A.T., 2015. DRAC: Dose Rate and Age Calculator for trapped charge dating. Quat. Geochronol. 28, 54–61.

Duller, G. (2015). Luminescence Dating. In: Rink, W.J., Thompson, J.W. (Eds.), Encyclopedia of Scientific Dating Methods, 390–404.

138

Edwards, M.B., 1979. Late Precambrian glacial loessites from North Norway and Svalbard. J Sediment Petrol 49, 85-92.

Ehlers, J., Gibbard, P.L., 2007. The extent and chronology of Cenozoic Global Glaciation. Quat. Int., From the Swiss Alps to the Crimean Mountains - Alpine Quaternary stratigraphy in a European context 164–165, 6–20.

Einwögerer, T., Friesinger, H., Händel, M., Neugebauer-Maresch, C., Simon, U., Teschler- Nicola, M., 2006. Upper Palaeolithic infant burials. Nature 444, 285.

Einwögerer, T., Händel, M., Neugebauer-Maresch, C., Simon, U., Steier, P., Teschler-Nicola, M., Wild, E.M., 2009. 14C dating of the Upper Paleolithic site at Krems-Wachtberg, Austria. Radiocarbon 51, 847-855.

Evans, M.E., 2001. Magnetoclimatology of aeolian sediments. Geophys. J. Int. 144: 495–497.

Evans, M.E., Heller, F., 2001. Magnetism of loess/palaeosol sequences: recent developments. Earth Sci. Rev. 54, 129–144.

Evans, M., Heller, F., 2003. Environmental Magnetism: Principles and Applications of Enviromagnetics. Academic press.

Evans, M.E., Rutter, N.W., Catto, N., Chlachula, J., Nyvlt, D., 2003. Magnetoclimatology: Teleconnection between the Siberian loess record and North Atlantic Heinrich events. Geology 31, 537–540.

Fedo, C.M., Nesbitt, H.W., Young, G.M., 1995. Unraveling the effects of potassium metasomatism in sedimentary rocks and paleosols, with implications for paleoweathering conditions and provenance. Geology 23, 921–924.

Feurdean, A., Perşoiu, A., Tanţău, I., Stevens, T., Magyari, E.K., Onac, B.P., Marković, S., Andrič, M., Connor, S., Fărcaş, S., Gałka, M., Gaudeny, T., Hoek, W., Kolaczek, P., Kuneš, P., Lamentowicz, M., Marinova, E., Michczyńska, D.J., Perşoiu, I., Płóciennik, M., Słowiński, M., Stancikaite, M., Sumegi, P., Svensson, A., Tămaş, T., Timar, A., Tonkov, S., Toth, M., Veski, S., Willis, K.J., Zernitskaya, V., 2014. Climate variability and associated vegetation response throughout Central and Eastern Europe (CEE) between 60 and 8 ka. Quat. Sci. Rev. 106, 206–224.

139

Fink, J., 1961. Die Gliederung des Jungpleistozäns in Oesterreich. Mitteilungen der Geologischen Gesellschaft in Wien 54, 1-25.

Fink, J., 1964. Die Subkommission für Lößstratigraphie der Internationalen Quartärvereinigung. Eiszeitalter und Gegenwart 15, 229-235.

Fink, J., Kukla, G.J., 1977. Pleistocene climates in central Europe; at least 17 interglacials after the Olduvai Event. Quaternary Research 7, 363-371.

Finlayson, C., Giles Pacheco, F., Rodríguez-Vidal, J., Fa, D.A., María Gutierrez López, J., Santiago Pérez, A., Finlayson, G., Allue, E., Baena Preysler, J., Cáceres, I., Carrión, J.S., Fernández Jalvo, Y., Gleed-Owen, C.P., Jimenez Espejo, F.J., López, P., Antonio López Sáez, J., Antonio Riquelme Cantal, J., Sánchez Marco, A., Giles Guzman, F., Brown, K., Fuentes, N., Valarino, C.A., Villalpando, A., Stringer, C.B., Martinez Ruiz, F., Sakamoto, T., 2006. Late survival of Neanderthals at the southernmost extreme of Europe. Nature 443, 850–853.

Fitzsimmons, K.E., Hambach, U., 2014. Loess accumulation during the last glacial maximum: Evidence from Urluia, southeastern Romania. Quat. Int. 334–335, 74–85.

Fitzsimmons, K.E., Hambach, U., Veres, D., Iovita, R., 2013. The Campanian Ignimbrite Eruption: New Data on Volcanic Ash Dispersal and Its Potential Impact on Human Evolution. PLoS ONE 8, e65839.

Fitzsimmons, K.E., Marković, S.B., Hambach, U., 2012. Pleistocene environmental dynamics recorded in the loess of the middle and lower Danube basin. Quat. Sci. Rev. 41, 104– 118.

Fleitmann, D., Cheng, H., Badertscher, S., Edwards, R.L., Mudelsee, M., Göktürk, O.M., Fankhauser, A., Pickering, R., Raible, C.C., Matter, A., Kramers, J., Tüysüz, O., 2009. Timing and climatic impact of Greenland interstadials recorded in stalagmites from northern Turkey. Geophys. Res. Lett. 36, L19707.

Folk, R.L., 1966. A review of grain-size parameters. Sedimentology 6, 73–93.

Forster, T., Evans, M.E., Heller, F., 1994. The frequency dependence of low field susceptibility in loess sediments. Geophys. J. Int. 118, 636–642.

140

Fu, Q., Hajdinjak, M., Moldovan, O.T., Constantin, S., Mallick, S., Skoglund, P., Patterson, N., Rohland, N., Lazaridis, I., Nickel, B., Viola, B., Prüfer, K., Meyer, M., Kelso, J., Reich, D., Pääbo, S., 2015. An early modern human from Romania with a recent Neanderthal ancestor. Nature 524, 216–219.

Fu, Q., Li, H., Moorjani, P., Jay, F., Slepchenko, S.M., Bondarev, A.A., Johnson, P.L.F., Aximu-Petri, A., Prüfer, K., de Filippo, C., Meyer, M., Zwyns, N., Salazar-García, D.C., Kuzmin, Y.V., Keates, S.G., Kosintsev, P.A., Razhev, D.I., Richards, M.P., Peristov, N.V., Lachmann, M., Douka, K., Higham, T.F.G., Slatkin, M., Hublin, J.-J., Reich, D., Kelso, J., Viola, T.B., Pääbo, S., 2014. Genome sequence of a 45,000-year- old modern human from western Siberia. Nature 514, 445–449.

Fuchs, M., Rousseau, D.-D., Antoine, P., Hatté, C., Gauthier, C., Marković, S., Zoeller, L., 2008. Chronology of the Last Climatic Cycle (Upper Pleistocene) of the Surduk loess sequence, Vojvodina, Serbia. Boreas 37, 66–73.

Galbraith, R.F., Roberts, R.G., Laslett, G.M., Yoshida, H., Olley, J.M., 1999a. Optical Dating of Single and Multiple Grains of Quartz from Jinmium Rock Shelter, Northern Australia: Part I, Experimental Design and Statistical Models*. Archaeometry 41, 339–364.

Galbraith, R.F., Roberts, R.G., Laslett, G.M., Yoshida, H., Olley, J.M., 1999b. Optical Dating of Single and Multiple Grains of Quartz from Jinmium Rock Shelter, Northern Australia: Part I, Experimental Design and Statistical Models*. Archaeometry 41, 339–364.

Gallet, S., Jahn, B., Torii, M., 1996. Geochemical characterization of the Luochuan loess paleosol sequence, China, and paleoclimatic implications. Chemical Geology 133, 67- 88.

Ganopolski, A., Winkelmann, R., Schellnhuber, H.J., 2016. Critical insolation–CO2 relation for diagnosing past and future glacial inception. Nature 529, 200–203.

Glazovskaya, M.A., 1952. Biological factors of weathering in high mountain regions (in Russian). Priroda 12, 106-110.

141

Goudie, A.S., Cooke, R.U., Doornkamp, J.C., 1979. The formation of silt from quartz dune sand by salt processes in deserts. J. Arid Environ. 2, 105-112.

Guerin, G., Mercier, N., Adamiec, G., 2011. Dose-rate conversion factors: update. Anc. TL 29, 5–8.

Guérin, G., Mercier, N., Nathan, R., Adamiec, G., Lefrais, Y., 2012. On the use of the infinite matrix assumption and associated concepts: A critical review. Radiat. Meas., Proceedings of the 13th International Conference on Luminescence and Electron Spin Resonance Dating, 10–14 July, 2011, Toruń, Poland 47, 778–785.

Guo, Z.T., Ruddiman, W.F., Hao, Q.Z., Wu, H.B., Qiao, Y.S., Zhu, R.X., Peng, S.Z., Wei, J.J., Yuan, B.Y., Liu, T.S., 2002. Onset of Asian desertification by 22 Myr ago inferred from loess deposits in China. Nature 416, 159–163.

Haase, D., Fink, J., Haase, G., Ruske, R., Pécsi, M., Richter, H., Altermann, M., Jager, K.D., 2007. Loess in Europe - its spatial distribution based on a Europeanloess map, scale 1:2,500,000. Quat. Sci. Rev. 26, 1301-1312.

Haesaerts, P., Damblon, F., Bachner, M., Trnka, G., 1996. Revised stratigraphy and chronology of the Willendorf II sequence, Lower Austria. Archaeologia Austriaca 80, 25-42.

Hahn, J., 1977. Aurignacien, das ältere Jungpaläolithikum in Mittel- und Osteuropa. Fundamenta: Reihe A, Bd. 9. Böhlau, Köln, Wien (in German).

Hao, Q., Guo, Z., Xu, B., Oldfield, F., 2010. Geochemical evidence for the provenance of middle Pleistocene loess deposits in southern China. Quat. Sci. Rev. 29, 3317–3326.

Hao, Q., Oldfield, F., Bloemendal, J., Guo, Z., 2008. The magnetic properties of loess and paleosol samples from the Chinese Loess Plateau spanning the last 22 million years. Palaeogeogr. Palaeoclimatol. Palaeoecol. 260, 389–404.

Hao, Q., Wang, L., Oldfield, F., Guo, Z., 2015. Extra-long interglacial in Northern Hemisphere during MISs 15-13 arising from limited extent of Arctic ice sheets in glacial MIS 14. Sci. Rep. 5, 12103.

142

Hao, Q., Wang, L., Oldfield, F., Peng, S., Qin, L., Song, Y., Xu, B., Qiao, Y., Bloemendal, J., Guo, Z., 2012. Delayed build-up of Arctic ice sheets during 400,000-year minima in insolation variability. Nature 490, 393–396.

Harnois, L., 1988. The CIW index: a new chemical index of weathering. Sed. Geol. 55, 319– 322.

Hatté, C., Gauthier, C., Rousseau, D.-D., Antoine, P., Fuchs, M., Lagroix, F., Marković, S.B., Moine, O., Sima, A., 2013. Excursions to C4 vegetation recorded in the Upper Pleistocene loess of Surduk (Northern Serbia): an organic isotope geochemistry study. Clim Past 9, 1001–1014.

Harzhauser, M., Piller, W.E., 2007. Benchmark data of a changing sea - Palaeogeography, Palaeobiogeography and events in the Central Paratethys during the Miocene. Palaeogeogr. Palaeoclimatol. Palaeoecol. 253, 8-31.

Heller, F., Liu, T., 1984. Magnetism of Chinese loess deposits. Geophys. J. Int. 77:125–141.

Heller, F., Liu, X., Liu, T., Xu, T., 1991.Magnetic susceptibility of loess in China. Earth Planet. Sci. Lett. 103:301–310.

Helmens, K.F., 2014. The Last Interglacial–Glacial cycle (MIS 5–2) re-examined based on long proxy records from central and northern Europe. Quat. Sci. Rev. 86, 115–143.

Higham, T., Compton, T., Stringer, C., Jacobi, R., Shapiro, B., Trinkaus, E., Chandler, B., Gröning, F., Collins, C., Hillson, S., O’Higgins, P., FitzGerald, C., Fagan, M., 2011. The earliest evidence for anatomically modern humans in northwestern Europe. Nature 479, 521–524.

Higham, T., Ramsey, C.B., Karavanić, I., Smith, F.H., Trinkaus, E., 2006. Revised direct radiocarbon dating of the Vindija G1 Upper Paleolithic Neandertals. Proc. Natl. Acad. Sci. U. S. A. 103, 553–557.

Hippolyte, J.-C., 2002. Geodynamics of Dobrogea (Romania): new constraints on the evolution of the Tornquist–Teisseyre Line, the Black Sea and the Carpathians. Tectonophysics, Paleostresses and Tectonics in the Peri-Tethyan Margins 357, 33–53.

143

Hrnjak, I., Lukić, T., Gavrilov, M.B., Marković , S.B., Unkašević, M., Tošić , I., 2014. Aridity in Vojvodina, Serbia. Theor. Appl. Climatol. 115, 323–332.

Hublin, J.-J., 2015. The modern human colonization of western Eurasia: when and where? Quat. Sci. Rev., Synchronising Environmental and Archaeological Records using Volcanic Ash Isochrons 118, 194–210.

Hughes, A.L.C., Gyllencreutz, R., Lohne, Ø.S., Mangerud, J., Svendsen, J.I., 2016. The last Eurasian ice sheets – a chronological database and time-slice reconstruction, DATED- 1. Boreas 45, 1–45.

Hughes, P.D., Woodward, J.C., Gibbard, P.L., 2007. Middle Pleistocene cold stage climates in the Mediterranean: New evidence from the glacial record. Earth Planet. Sci. Lett. 253, 50–56.

Hughes, P.D., Woodward, J.C., Gibbard, P.L., 2006. Late Pleistocene glaciers and climate in the Mediterranean. Glob. Planet. Change 50, 83–98.

Hughes, P.D., Woodward, J.C., van Calsteren, P.C., Thomas, L.E., 2011. The glacial history of the Dinaric Alps, Montenegro. Quat. Sci. Rev. 30, 3393–3412.

Huntley, D.J., Baril, M.R., 1997. The K content of the K-feldspars being measured in optical dating or in thermoluminescence dating. Anc. TL 15, 11–12.

Huntley, D.J., Hancock, R.G.V., 2001. The Rb contents of the K-feldspar grains being measured in optical dating. Anc. TL 19, 43–46.

Huntley, D.J., Lamothe, M., 2001. Ubiquity of anomalous fading in K-feldspars and the measurement and correction for it in optical dating. Can. J. Earth Sci. 38, 1093–1106.

Iovita, R., Fitzsimmons, K.E., Dobos, A., Hambach, U., Hilgers, A., Zander, A., 2012. Dealul Guran: evidence for Lower Palaeolithic (MIS 11) occupation of the Lower Danube loess steppe. Antiquity 86, 973-989.

Iriondo, M.H., Kröhling, D.M., 2007. Non-classical types of loess. Sed. Geol. 202, 352-368.

ISO 10693, 1995. Soil Quality – Determination of Carbonate Content – Volumetric Method, Geneva.

144

ISO 13320-1, 1999. Particle Size Analysis – Laser Diffraction Methods. Part 1. General principles, Annex A: Theoretical Background of Laser Diffraction, Geneva.

ISO 14688-1, 2002. Geotechnical Engineering – Identification and Classification of Soil. Part 1. Identification and Description, Geneva.

Ivanov, D., Utescher, T., Mosbrugger, V., Syabryaj, S., Djordjevic-Milutinovic, D., Molchanoff, S., 2011. Miocene vegetation and climate dynamics in Eastern and Central Paratethys (Southeastern Europe). Palaeogeogr. Palaeoclimatol. Palaeoecol. 304, 262-275.

Jary, Z., Ciszek, D., 2013. Late Pleistocene loess–palaeosol sequences in Poland and western Ukraine. Quat. Int. 296, 37–50.

Jeong, G.Y., Hillier, S., Kemp, R.A., 2008. Quantitative bulk and single particle mineralogy of a thick Chinese loess–paleosol section: implications for loess provenance and weathering. Quat. Int. 27, 1271–1287.

Ji, J., Balsam, W., Chen, J., Liu, L., 2002. Rapid and quantitative measurement of hematite and goethite in the Chinese loessepaleosol sequence by diffuse reflectance spectroscopy. Clays Clay Miner. 50 (2), 208-216.

Jipa, D.C., 2014. The conceptual sedimentary model of the Lower Danube loess basin: Sedimentogenetic implications. Quat. Int., Loess and the record of upper Palaeolithic cultures in the Danube Basin 351, 14–24.

Jordanova, D., Petersen, N., 1999. Palaeoclimatic record from a loess-soil profile in northeastern Bulgaria - I. Rockmagnetic properties. Geophys. J. Int. 138, 520–532.

Kabata-Pendias, A., 2010. Trace Elements in Soils and Plants, Fourth Edition. CRC Press.

Kels, H., Protze, J., Sitlivy, V., Hilgers, A., Zander, A., Anghelinu, M., Bertrams, M., Lehmkuhl, F., 2014. Genesis of loess-like sediments and soils at the foothills of the Banat Mountains, Romania – Examples from the Paleolithic sites Româneşti and Coşava. Quat. Int. 351, 213-230.

Koch, R., Neumeister, H., 2005. Zur Klassifikation von Lößsedimenten nach genetischen Kriterien. Zeitschrift für Geomorphologie NF 49, 183-203.

145

Kohfeld, K.E., Harrison, S.P., 2001. DIRTMAP: the geological record of dust. Earth Sci. Rev. 54, 81-114.

Konert, M., Vandenberghe, J., 1997. Comparison of laser grain size analysis with pipette and sieve analysis: a solution for the under-estimation of the clay fraction. Sedimentology 44, 523–535.

Kostić, N., Protić, N., 2000. Pedology and mineralogy of loess profiles at Kapela-Batajnica and Stalać, Serbia. CATENA 41, 217–227.

Krauß, L., Zens, J., Zeeden, C., Schulte, P., Eckmeier, E., and Lehmkuhl, F. (2016). A Multi- Proxy Analysis of two Loess-Paleosol Sequences in the Northern Harz Foreland, Germany. Palaeogeogr. Palaeoclimatol. Palaeoecol. 461, 401–417.

Kreutzer, S., Schmidt, C., DeWitt, R., Fuchs, M., 2014. The a-value of polymineral fine grain samples measured with the post-IR IRSL protocol. Radiat. Meas. 69, 18–29.

Kronberg, B.I., Nesbitt, H.W., 1981. Quantification of weathering, soil geochemistry and soil fertility. J. Soil Sci. 32, 453–459.

Kukla, G.J., 1977. Pleistocene land-sea correlations I. Europe. Earth-Sci. Rev. 13, 307-374.

Kukla, G.J., 1987. Loess stratigraphy in Central China. Quat. Sci. Rev. 6, 191-219.

Kukla, G.J., An, Z.S., 1989. Loess stratigraphy in Central China. Palaeogeogr. Palaeoclimatol. Palaeoecol. 72, 203-225.

Kukla, G.J., Heller, F., Liu, X.M., Xu, T.C., Liu, T.S., An, Z.S., 1988. Pleistocene climates in China dated by magnetic susceptibility. Geology 16, 811-814.

Lehmkuhl, F., 1997. The spatial distribution of loess and loess-like sediments in the mountain areas of Central and High Asia. Zeitschrift fur Geomorphologie 111, 97–116.

Lehmkuhl, F., Schulte, P., Zhao, H., Hülle, D., Protze, J., Stauch, G., 2014. Timing and spatial distribution of loess and loess-like sediments in the mountain areas of the northeastern Tibetan Plateau. CATENA 117, 23-33.

Leever, K.A., Matenco, L., Garcia-Castellanos, D., Cloetingh, S.A.P.L., 2011. The evolution of the Danube gateway between Central and Eastern Paratethys (SE Europe): insight

146

from numerical modelling of the causes and effects of connectivity between basins and its expression in the sedimentary record. Tectonophysics 502, 175-195.

Leicher, N., Zanchetta, G., Sulpizio, R., Giaccio, B., Wagner, B., Nomade, S., Francke, A., Del Carlo, P., 2015. First tephrostratigraphic results of the DEEP site record from Lake Ohrid, Macedonia. Biogeosciences Discuss. 12, 15411–15460.

Leng, M.J., Wagner, B., Boehm, A., Panagiotopoulos, K., Vane, C.H., Snelling, A., Haidon, C., Woodley, E., Vogel, H., Zanchetta, G., Baneschi, I., 2013. Understanding past climatic and hydrological variability in the Mediterranean from Lake Prespa sediment isotope and geochemical record over the Last Glacial cycle. Quat. Sci. Rev., International Association of Limnogeology – Isotopes and Lakes 66, 123–136.

Leonhard, K.C.v., 1824. Charakteristik der Felsarten - Dritte Abteilung. Trümmer-Gesteine. Lose Gesteine. Kohlen. Joseph Engelmann, Heidelberg.

Liang, Y., Yang, T., Velichko, A.A., Zeng, B., Shi, P., Wang, L., He, Y., Chen, J., Chen, Y., 2016. Paleoclimatic record from Chumbur-Kosa section in Sea of Azov region since Marine Isotope Stage 11. J. Mt. Sci. 13, 985–999.

Liritzis, I., Stamoulis, K., Papachristodoulou, C., Ioannides, K., 2013. A re-evaluation of radiation dose-rate conversion factors. Mediterr Archaeol Archaeom 13, 1–15.

Lisiecki, L.E., Raymo, M.E., 2005. A Pliocene-Pleistocene stack of 57 globally distributed benthic δ18O records. Paleoceanography 20, PA1003.

Liu, T.S., 1988. Loess in China, second ed. Springer-Verlag, Berlin & Heidelberg.

Liu, Q., M. J. Jackson, S. K. Banerjee, B. A. Maher, C. Deng, Y. Pan, and R. Zhu (2004), Mechanism of the magnetic susceptibility enhancements of the Chinese loess, J. Geophys. Res. 109, B12107.

Liu, X.M., Liu, Z., Lü, B., Marković , S.B., Chen, J.S., Guo, H., Ma, M.M., Zhao, G.Y., Feng, H., 2013. The magnetic properties of Serbian loess and its environmental significance. Chin. Sci. Bull. 58, 353–363.

Lowe, J., Barton, N., Blockley, S., Ramsey, C.B., Cullen, V.L., Davies, W., Gamble, C., Grant, K., Hardiman, M., Housley, R., Lane, C.S., Lee, S., Lewis, M., MacLeod, A.,

147

Menzies, M., Müller, W., Pollard, M., Price, C., Roberts, A.P., Rohling, E.J., Satow, C., Smith, V.C., Stringer, C.B., Tomlinson, E.L., White, D., Albert, P., Arienzo, I., Barker, G., Borić, D., Carandente, A., Civetta, L., Ferrier, C., Guadelli, J.-L., Karkanas, P., Koumouzelis, M., Müller, U.C., Orsi, G., Pross, J., Rosi, M., Shalamanov-Korobar, L., Sirakov, N., Tzedakis, P.C., 2012. Volcanic ash layers illuminate the resilience of Neanderthals and early modern humans to natural hazards. Proc. Natl. Acad. Sci. 109, 13532–13537.

Ložek, V., 1965. Das Problem der Lößbildung und die Lößmollusken. Eiszeitalter und Gegenwart 16, 61-75.

Luehmann, M.D., Schaetzl, R.J., Miller, B.A., Bigsby, M.E., 2013. Thin, pedoturbated, and locally sourced loess in the western Upper Peninsula of Michigan. Aeolian Res. 8, 85– 100.

Lukić, T., Basarin, B., Buggle, B., Markovič, S.B., Tomović, V.M., Raljič, J.P., Hrnjak, I., Timar-Gabor, A., Hambach, U., Gavrilov, M.B., 2014. A joined rock magnetic and colorimetric perspective on the Late Pleistocene climate of Orlovat loess site (Northern Serbia). Quat. Int. 334–335, 179–188.

Lyell, C., 1834. Observations on the loamy deposit called “loess” of the Basin of the Rhine. Edinburgh New Philosophical Journal 17, 110-113.

Maher, B.A., 2011. The magnetic properties of quaternary aeolian dusts and sediments, and their palaeoclimatic significance. Aeolian Res. 3:87–144. Makeev, A.O., 2009. Pedogenic alteration of aeolian sediments in the upper loess mantles of the Russian Plain. Quat. Int. 209, 79-94.

Marković, S., Svirčev, Z., Tomić, P., Bogdanović, Ž, Lazić, L, Jovanović, M, Pavić, D., 1998. Pokazatelji i uzroci smanjenja kvaliteta vode u jugoslovenskom delu toka reke Tamiš. In: Naš Tamiš, Marković, S. and Svirčev, Z. (Eds.), Institute for Geography, Novi Sad, 11-18.

Marković, S.B., Svirčev, Z. (eds.) (1998): Our Tamiš. Scientific monography. Institute of Geography, FNS, Novi Sad, 172.

148

Marković, S.B., Kostić, N., Oches, E.A., 2004. Paleosols in the Ruma loess section. Revista Mexicana de Ciencias Geológicas 21, 79–87.

Marković , S.B., McCoy, W.D., Oches, E.A., Savić , S., Gaudenyi, T., Jovanović, M., Stevens, T., Walther, R., Ivanišević, P., Galić, Z., 2005. Paleoclimate record in the Late Pleistocene loess–paleosol sequence at Petrovaradin Brickyard (Vojvodina, Serbia). Geol. Carpath. 56, 483–491.

Marković, S.B., Hambach, U., Stevens, T., Jovanović, M., O’Hara-Dhand, K., Basarin, B., Lu, H., Smalley, I., Buggle, B., Zech, M., Svirčev, Z., Sümegi, P., Milojković, N., Zöller, L., 2012a. Loess in the Vojvodina region (Northern Serbia): an essential link between European and Asian Pleistocene environments. Neth. J. Geosci. 91, 173–188.

Marković, S.B., Hambach, U., Stevens, T., Basarin, B., O’Hara-Dhand, K., Gavrilov, M.M., Gavrilov, M.B., Smalley, I., Teofanov, N., 2012b. Relating the Astronomical Timescale to the Loess–Paleosol Sequences in Vojvodina, Northern Serbia, in: Berger, A., Mesinger, F., Sijacki, D. (Eds.), Climate Change. Springer Vienna, pp. 65–78.

Marković, S.B., Bokhorst, M.P., Vandenberghe, J., McCoy, W.D., Oches, E.A., Hambach, U., Gaudenyi, T., Jovanović, M., Zöller, L., Stevens, T., Machalett, B., 2008. Late Pleistocene loess-palaeosol sequences in the Vojvodina region, north Serbia. J. Quat. Sci. 23, 73–84.

Marković, S.B., Hambach, U., Stevens, T., Kukla, G.J., Heller, F., McCoy, W.D., Oches, E.A., Buggle, B., Zöller, L., 2011. The last million years recorded at the Stari Slankamen (Northern Serbia) loess-palaeosol sequence: revised chronostratigraphy and long-term environmental trends. Quat. Sci. Rev. 30, 1142–1154.

Marković, S.B., Oches, E., Sümegi, P., Jovanović, M., Gaudenyi, T., 2006. An introduction to the Middle and Upper Pleistocene loess–paleosol sequence at Ruma brickyard, Vojvodina, Serbia. Quat. Int. 149, 80–86.

Marković, S.B., Oches, E.A., McCoy, W.D., Frechen, M., Gaudenyi, T., 2007. Malacological and sedimentological evidence for “warm” glacial climate from the Irig loess sequence, Vojvodina, Serbia. Geochem. Geophys. Geosystems 8, Q09008.

149

Marković, S.B., Hambach, U., Catto, N., Jovanović, M., Buggle, B., Machalett, B., Zöller, L., Glaser, B., Frechen, M., 2009. The Middle and Late Pleistocene loess–paleosol sequences at Batajanica, Vojvodina, Serbia. Quat. Int. 198, 255–266.

Marković, S.B., Stevens, T., Kukla, G.J., Hambach, U., Fitzsimmons, K.E., Gibbard, P., Buggle, B., Zech, M., Guo, Z., Hao, Q., Wu, H., O’Hara Dhand, K., Smalley, I.J., Újvári, G., Sümegi, P., Timar-Gabor, A., Veres, D., Sirocko, F., Vasiljević, D.A., Jary, Z., Svensson, A., Jović, V., Lehmkuhl, F., Kovács, J., Svirčev, Z., 2015. Danube loess stratigraphy — Towards a pan-European loess stratigraphic model. Earth-Sci. Rev. 148, 228–258.

Marković, S.B., Timar-Gabor, A., Stevens, T., Hambach, U., Popov, D., Tomić, N., Obreht, I., Jovanović, M., Lehmkuhl, F., Kels, H., Marković, R., Gavrilov, M.B., 2014a. Environmental dynamics and luminescence chronology from the Orlovat loess– palaeosol sequence (Vojvodina, northern Serbia). J. Quat. Sci. 29, 189–199.

Marković, S.B., Korać, M., Mrdjiá, N., Buylaert, J.P., Thiel, C., McLaren, S.J., Stevens, T., Tomić, N., Petić, N., Jovanović, M., Vasiljević, Dj. A., Sümegi, P., Gavrilov, M.B., Obreht, I., 2014b. Palaeoenvironment and geoconservation of mammoths from the Nosak loess–palaeosol sequence (Drmno, northeastern Serbia): initial results and perspectives. Quat. Int. 334–335, 30–39.

Marti, A., Folch, A., Costa, A., Engwell, S., 2016. Reconstructing the plinian and co- ignimbrite sources of large volcanic eruptions: A novel approach for the Campanian Ignimbrite. Sci. Rep. 6, 21220.

McLennan, S.M., Hemming, S., McDaniel, D.K., Hanson, G.N., 1993. Geochemical approaches to sedimentation, provenance and tectonics. In: Johnsson, M.J., Basu, A. (Eds.), Processes Controlling the Composition of Clastic Sediments, vol 284. Geological Society of America Special Paper, pp. 21–40.

Mejdahl, V., 1979. Thermoluminescence Dating: Beta-Dose Attenuation in Quartz Grains. Archaeometry 21, 61–72.

Mellars, P., 2006. A new radiocarbon revolution and the dispersal of modern humans in Eurasia. Nature 439, 931–935.

150

Ménot, G., Bard, E., 2012. A precise search for drastic temperature shifts of the past 40,000 years in southeastern Europe. Paleoceanography 27, PA2210.

Morley, M.W., Woodward, J.C., 2011. The Campanian Ignimbrite (Y5) tephra at Crvena Stijena Rockshelter, Montenegro. Quat. Res. 75, 683–696.

Muhs, D.R., Bettis, E.A., Been, J., McGeehin, J.P., 2001. Impact of climate and parent material on chemical weathering in loess-derived soils of the Mississippi River valley. Soil Sci. Soc. Am. J. 65 (6), 1761-1777.

Muhs, D.R., Ager, T.A., Bettis III, E.A., McGeehin, J., Been, J.M., Begét, J.E., Pavich, M.J., Stafford Jr., T.W., de Stevens, A.S.P., 2003. Stratigraphy and palaeoclimatic significance of Late Quaternary loessepalaeosol sequences of the Last InterglacialeGlacial cycle in central Alaska. Qua. Sci. Rev.22, 1947-1986.

Muhs, D.R., Bettis III, E.A., 2003. Quaternary loess-Paleosol sequences as examples of climate-driven sedimentary extremes. The Geological Society of America Special Papers 370, 53-74.

Muhs, D.R., Benedict, J.B., 2006. Eolian additions to late Quaternary alpine soils, Indian Peaks Wilderness area, Colorado Front Range. Arct. Antarct. Alp. Res. 38, 120–130.

Muhs, D.R., Bettis III, E.A., Aleinikoff, J.N., McGeehin, J.P., Beann, J., Skipp, G., Marshall, B.D., Roberts, H.M., Johnson, W.C., Benton, R., 2008. Origin and paleoclimatic significance of late Quaternary loess in Nebraska: Evidence from stratigraphy, chronology, sedimentology, and geochemistry. GSA Bull. 120, 13781407.

Muhs, D.R., Cattle, S.R., Crouvi, O., Rousseau, D.-D., Sun, J.M., Zárate, M.A., 2014. Loess records. In: Knippertz, P., Stuut, J.-B.W. (Eds.), Mineral Dust. Springer Netherlands, Dordrecht, pp. 411-441.

Murray, A.S. and Wintle, A.G., 2000. Luminescence dating of quartz using an improved single-aliquot regenerative-dose protocol. Radiat. Meas. 32, 57-73.

Murray, A.S. and Wintle, A.G., 2003. The single aliquot regenerative dose protocol: potential for improvements in reliability. Radiat. Meas. 37, 377-381.

151

Necula, C., Dimofte, D., Panaiotu, C., 2015. Rock magnetism of a loess-palaeosol sequence from the western Black Sea shore (Romania). Geophys. J. Int. 202, 1733–1748.

Necula, C., Panaiotu, C., 2008. Application of dynamic programming to the dating of a loess- paleosol sequence. Romanian Rep. Phys. 60, 157–171.

Necula, C., Panaiotu, C., Heslop, D., Dimofte, D., 2013. Climatic control of magnetic granulometry in the Mircea Vodă loess/paleosol sequence (Dobrogea, Romania). Quat. Int., Advancing Pleistocene and Holocene climate change research in the Carpathian- Balkan region 293, 5–14.

Nesbitt, H.W., Markovics, G., Price, R.C., 1980. Chemical processes affecting alkalis and alkaline earths during continental weathering. Geochim. Cosmochim. Acta 44, 1659- 1666.

Nesbitt, H.W., Young, G.M., 1982. Early Proterozoic climates and plate motions inferred from major element chemistry of lutites. Nature 299, 715–717.

Nesbitt, H.W., Young, G.M., 1984. Prediction of some weathering trends of plutonic and volcanic rocks based on thermodynamic and kinetic considerations. Geochim. Cosmochim. Acta 48, 1523–1534.

Nesbitt, H.W., Young, G.M., McLennan, S.M., Keays, R.R., 1996. Effects of chemical weathering and sorting on the petrogenesis of siliciclastic sediments, with implications for provenance studies. J. Geol. 104, 525–542.

Nesbitt, H.W., Young, G.M., 1989. Formation and diagenesis of weathering profiles. J. Geol. 97, 129–147.

Nie, J., Stevens, T., Rittner, M., Stockli, D., Garzanti, E., Limonta, M., Bird, A., Andò, S., Vermeesch, P., Saylor, J., Lu, H., Breecker, D., Hu, X., Liu, S., Resentini, A., Vezzoli, G., Peng, W., Carter, A., Ji, S., Pan, B., 2015. Loess Plateau storage of Northeastern Tibetan Plateau-derived Yellow River sediment. Nat. Commun. 6, 8511.

Nigst, P.R., Haesaerts, P., Damblon, F., Frank-Fellner, C., Mallol, C., Viola, B., Götzinger, M., Niven, L., Trnka, G., Hublin, J.-J., 2014. Early modern human settlement of

152

Europe north of the Alps occurred 43,500 years ago in a cold steppe-type environment. Proc. Natl. Acad. Sci. 201412201.

Nottebaum, V., Stauch, G., Hartmann, K., Zhang, J., and Lehmkuhl, F. (2015). Unmixed loess grain size populations along the northern Qilian Shan (China): Relationships between geomorphologic, sedimentologic and climatic controls. Quat. Int. 372, 151–166.

Obreht, I., Buggle, B., Catto, N., Markovič, S.B., Bösel, S., Vandenberghe, D.A.G., Hambach, U., Svirčev, Z., Lehmkuhl, F., Basarin, B., Gavrilov, M.B., Jović, G., 2014. The Late Pleistocene Belotinac section (southern Serbia) at the southern limit of the European loess belt: Environmental and climate reconstruction using grain size and stable C and N isotopes. Quat. Int., 334–335, 10–19.

Obreht, I., Zeeden, C., Hambach, U., Veres, D., Marković, S.B., Bösken, J., Svirčev, Z., Bačević, N., Gavrilov, M.B., Lehmkuhl, F., 2016. Tracing the influence of Mediterranean climate on Southeastern Europe during the past 350,000 years. Sci. Rep. 6, 36334.

Obreht, I., Zeeden, C., Schulte, P., Hambach, U., Eckmeier, E., Timar-Gabor, A., Lehmkuhl, F., 2015. Aeolian dynamics at the Orlovat loess–paleosol sequence, northern Serbia, based on detailed textural and geochemical evidence. Aeolian Res. 18, 69–81.

Obruchev, V.A., 1945. Loess types and their origin. Am. J. Sci. 243, 256-262.

Özer, M., Orhan, M., Işik, N.S., 2010. Effect of particle optical properties on size distribution of soils obtained by laser diffraction. Environ. Eng. Geosci. 16, 163–173.

Panagiotopoulos, K., Böhm, A., Leng, M.J., Wagner, B., Schäbitz, F., 2014. Climate variability over the last 92 ka in SW Balkans from analysis of sediments from Lake Prespa. Clim Past 10, 643–660.

Pavić, D. (2006): Hydrological characteristics of the Tisa river in Serbia and posibilities for nautic tourism development. Ph. D. thesis. Department of Geography, Tourism, and Hotel Management, Faculty of Natural Sciences. pp. 343.

Pécsi, M., 1990. Loess is not just the accumulation of dust. Quat. Int. 7, 1–21.

153

Pécsi, M., 1994. A landslide type occurring frequently along the loess bluff in the Hungarian Danube section. Quat. Int. 24, 31-33.

Pécsi, M., 1995. The role of principles and methods in loess-paleosol investigations. GeoJournal 36, 117-131.

Pécsi, M., Richter, G., 1996. Löss: Herkunft - Gliederung - Landschaften. Zeitschrift für Geomorphologie N.F. Supplementband 98. Gebrüder Borntrger, Berlin & Stuttgart.

Pekarova, P., Pekar, J. 2006. Long-term discharge prediction for the Turnu Severin station (the Danube) using a linear autoregressive model. Hydrol. Process. 20, 1217–1228.

Popov, D., Marković, S.B., Štrbac, D., 2008. Generations of meanders in Serbian part of Tisa valley. Zbornik radova Geografskog instituta ‘‘Jovan Cvijic´ ’’ SANU 58, 31–42.

Popov, D., Marković, S.B., Jovanović, M., Mesaroš, Arsenović, D., Stankov, U., Gubik, D., 2012. Geomorphological investigations and GIS approach of the Tamiš loess plateau, Banat region (Northern Serbia). Geogr. Pannon. 16, 1–9.

Posea, G. 2002. Geografia României, Manual pentru clasa a XII-a, Grigore Posea, 67-69.

Prescott, J.R., Hutton, J.T., 1994. Cosmic ray contributions to dose rates for luminescence and ESR dating: Large depths and long-term time variations. Radiat. Meas. 23, 497–500.

Preusser, F., Geyh, M.A. and Schlüchter, C., 2003. Timing of Late Pleistocene climate change in lowland Switzerland. Quat. Sci. Rev. 22, 1435-1445.

Prins, M.A., Vriend, M.G.A., Nugteren, G., Vandenberghe, J., Lu, H., Zheng, H., Weltje, G.J., 2007. Late quaternary aeolian dust flux variability on the Chinese loess Plateau: inferences from unmixing of loess grain-size records. Quat. Sci. Rev.26, 242–254.

Prohaska, S. (1998): The main hydrological characteristcs of Yugoslav part of Tamiš river. In: Our Tamiš. Marković, S. B. i Svirčev, Z. (eds.). Institute of Geography, Faculty of Natural Sciences, Novi Sad: 19-36.

Pye, K., 1987. Aeolian Dust and Dust Deposits. Academic Press, London.

Pye, K., 1995. The nature, origin and accumulation of loess. Quat. Sci. Rev., Aeolian Sediments in the Quaternary Record 14, 653–667.

154

Pye, K., Blott, S.J., 2004. Particle size analysis of sediments, soils and related particulate materials for forensic purposes using laser granulometry. Forensic Sci. Int. 144, 19– 27.

Pye, K., Sherwin, D., 1999. Loess. In: Goudie, A.S., Livingstone, I., Stokes, S. (Eds.), Aeolian Environments, Sediments and Landforms. Wiley & Sons, pp. 213-238

Pyle, D.M., Ricketts, G.D., Margari, V., van Andel, T.H., Sinitsyn, A.A., Praslov, N.D., Lisitsyn, S., 2006. Wide dispersal and deposition of distal tephra during the Pleistocene “Campanian Ignimbrite/Y5” eruption, Italy. Quat. Sci. Rev. 25, 2713– 2728.

Republic Hydrometeorological Service of Serbia. Available at: http://www.hidmet.gov.rs/eng/meteorologija/stanica_sr.php?moss_id=13383 (Accessed: 15th January 2016).

Richthofen, E.V., 1882. On the mode of origin of the loess. Geological Magazine 9, 293-305.

Rohling, E.J., Foster, G.L., Grant, K.M., Marino, G., Roberts, A.P., Tamisiea, M.E., Williams, F., 2014. Sea-level and deep-sea-temperature variability over the past 5.3 million years. Nature 508, 477–482.

Rohling, E.J., Medina-Elizalde, M., Shepherd, J.G., Siddall, M., Stanford, J.D., 2012. Sea Surface and High-Latitude Temperature Sensitivity to Radiative Forcing of Climate over Several Glacial Cycles. J. Clim. 25, 1635–1656.

Rousseau, D.D., Antoine, P., Hatté, C., Lang, A., Zöller, L., Fontugne, M., Ben Othman, D., Luck, J.M., Moine, O., Labonne, M., Bentaleb, I., Jolly, D., 2002. Abrupt millennial climatic changes from Nussloch (Germany) Upper Weichselian eolian records during the Last Glaciation. Quat. Sc. Revi.21, 1577-1582.

Russell, R.J., 1944. Lower Mississippi valley loess. Geological Society of America Bulletin 55, 1-40.

Sadori, L., Koutsodendris, A., Panagiotopoulos, K., Masi, A., Bertini, A., Combourieu- Nebout, N., Francke, A., Kouli, K., Joannin, S., Mercuri, A.M., Peyron, O., Torri, P., Wagner, B., Zanchetta, G., Sinopoli, G., Donders, T.H., 2016. Pollen-based

155

paleoenvironmental and paleoclimatic change at Lake Ohrid (south-eastern Europe) during the past 500 ka. Biogeosciences 13, 1423–1437.

Sarala, P., Väliranta, M., Eskola, T., Vaikutiené, G., 2016. First physical evidence for forested environment in the Arctic during MIS 3. Sci. Rep. 6, 29054.

Scaillet, S., Vita-Scaillet, G., Rotolo, S.G., 2013. Millennial-scale phase relationships between ice-core and Mediterranean marine records: insights from high-precision 40Ar/39Ar dating of the Green Tuff of Pantelleria, Sicily Strait. Quat. Sci. Rev. 78, 141–154.

Schaller, K., 2000. Praktikum zur Bodenkunde und Pflanzenernährung, Geisenheimer Berichte (2), Geisenheim, 8th ed., 586 p. (in German).

Schaetzl, R.J., Attig, J.W., 2013. The loess cover of northeastern Wisconsin. Quat. Res. 79, 199–214.

Schatz, A.K., Zech, M., Buggle, B., Gulyás, S., Hambach, U., Marković, S.B., Sümegi, P., Scholten, T., 2011. The late Quaternary loess record of Tokaj, Hungary: reconstructing palaeoenvironment, vegetation and climate using stable C and N isotopes and biomarkers. Quat. Int. 240, 52–61.

Schatz, A.-K., Scholten, T., Kühn, P., 2015. Paleoclimate and weathering of the Tokaj (Hungary) loess–paleosol sequence. Palaeogeogr. Palaeoclimatol. Palaeoecol. 426, 170–182.

Schmidt, E.D., Machalett, B., Marković, S.B., Tsukamoto, S., Frechen, M., 2010. Luminescence chronology of the upper part of the Stari Slankamen loess sequence (Vojvodina, Serbia). Quat. Geochronol. 5, 137–142.

Schulte, P., Lehmkuhl, F., Steininger, F., Loibl, D., Lockot, G., Protze, J., Fischer, P., Stauch, G., 2016. Influence of HCl pretreatment and organo-mineral complexes on laser diffraction measurement of loess–paleosol-sequences. CATENA 137, 392–405.

Sebe, K., Csillag, G., Ruszkiczay-Ruediger, Z., Fodor, L., Thamo-Bozso, E., Mueller, P., Braucher, R., 2011. Wind erosion under cold climate: a Pleistocene periglacial mega-

156

yardang system in Central Europe (Western Pannonian Basin, Hungary). Geomorphology 134, 470–482.

Semmel, A., 1968. Studien über den Verlauf jungpleistozäner Formung in Hessen. Frankfurter Geographische Hefte 45. Kramer, Frankfurt am Main.

Semmel, A., 1993. Grundzüge der Bodengeographie, third ed. Teubner, Stuttgart.

Shatravin, V., 2007. Reconstruction of Pleistocene and Holocene glaciations in the Tien-Shan and Pamir from new starting positions (in Russian). In: Climate, Glaciers and Lakes of the Tien Shan: Journey to the Past, Bishkek, pp. 26-46.

Sheldon, N.D., Tabor, N.J., 2009. Quantitative paleoenvironmental and paleoclimatic reconstruction using paleosols. Earth Sci. Rev. 95, 1-52.

Shumilovskikh, L.S., Fleitmann, D., Nowaczyk, N.R., Behling, H., Marret, F., Wegwerth, A., Arz, H.W., 2014. Orbital- and millennial-scale environmental changes between 64 and 20 ka BP recorded in Black Sea sediments. Clim Past 10, 939–954.

Signorelli, S., Vaggelli, G., Francalanci, L., Rosi, M., 1999. Origin of magmas feeding the Plinian phase of the Campanian Ignimbrite eruption, Phlegrean Fields (Italy): constraints based on matrix-glass and glass-inclusion compositions. J. Volcanol. Geotherm. Res. 91, 199–220.

Sinitsyn, A.A., Hoffecker, J.F., 2006. Radiocarbon dating and chronology of the Early Upper Paleolithic at Kostenki. Quaternary International 152-153, 164-174.

Sirocko, F., Knapp, H., Dreher, F., Förster, M.W., Albert, J., Brunck, H., Veres, D., Dietrich, S., Zech, M., Hambach, U., Röhner, M., Rudert, S., Schwibus, K., Adams, C., Sigl, P., 2016. The ELSA-Vegetation-Stack: Reconstruction of Landscape Evolution Zones (LEZ) from laminated Eifel maar sediments of the last 60,000 years. Glob. Planet. Change 142, 108–135.

Sitlivy, V., Chabai, V., Anghelinu, M., Uthmeier, T., Kels, H., Hilgers, A., Schmidt, C., Nita, L., Baltean, I., Veselsky, A., Hauck, T., 2012. The earliest Aurignacian in Romania: new investigations at the open air site of Românești-Dumbravita I (Banat). Quartär 59, 85–130.

157

Smalley, I.J., 1966. Properties of glacial loess and formation of loess deposits. J. Sediment. Petrol. 36, 669-676.

Smalley, I.J., 1971. In-situ theories of loess formation and significance of calciumcarbonate content of loess. Earth Sci. Rev. 7, 67-85.

Smalley, I.J., Vita-Finzi, C., 1968. Formation of Fine particles in Sandy deserts and nature of desert loess. J. Sediment. Petrol. 38, 766-774.

Smalley, I.J., Krinsley, D.H., 1978. Loess deposits associated with deserts. CATENA 5, 53- 66.

Smalley, I.J., Leach, J.A., 1978. The origin and distribution of the loess in the Danube basin and associated regions of East-Central Europe - a review. Sediment. Geol. 21, 1–26.

Smalley, I.J., 1990. Possible formation mechanisms for the modal coarse-silt quartz particles in loess deposits. Quat. Int.l 7-8, 23-27.

Smalley, I.J., 1995. Making the material: the formation of silt-sized primary mineral particles for loess deposits. Quat. Sci. Rev. 14, 645-651.

Smalley, I.J., Jefferson, I.F., Dijkstra, T.A., Derbyshire, E., 2001. Some major events in the development of the scientific study of loess. Earth Sci. Rev. 54, 5-18.

Smalley, I.J., O'Hara-Dhand, K., Wint, J., Machalett, B., Jary, Z., Jefferson, I., 2009. Rivers and loess: the significance of long river transportation in the complex event-sequence approach to loess deposit formation. Quat. Int. 198, 7-18.

Smalley, I.J., Marković, S.B., O'Hara Dhand, K., Wynn, P., 2010. A man from Bendery: L. S. Berg as geographer and loess scholar. Geologos 16, 111-119.

Smalley, I., Marković, S.B., Svirčev, Z., 2011. Loess is [almost totally formed by] the accumulation of dust. Quat. Int. 240, 4–11.

Smalley, I.J., Marković, S.B., 2014. Loessification and hydroconsolidation: there is a connection. CATENA 117, 94-99.

158

Soreghan, G.S., Soreghan, M.J., Hamilton, M.A., 2008. Origin and significance of loess in late Paleozoic western Pangaea: a record of tropical cold? Palaeogeogr. Palaeoclimatol. Palaeoecol. 268, 234-259.

Spooner, N.A., 1994. The anomalous fading of infrared-stimulated lumi-nescence from feldspars. Radiat. Meas. 23(2-3): 625-632.

Sprafke, T., Terhorst, B., Peticzka, R., Thiel, C., 2013. Paudorf locus typicus (Lower Austria) revisited - the potential of the classic loess outcrop for Middle to Late Pleistocene landscape reconstructions. Quat. Sci. J. (E&G) 62, 59-72.

Sprafke, T., Thiel, C., Terhorst, B., 2014. From micromorphology to palaeoenvironment: the MIS 10 to MIS 5 record in Paudorf (Lower Austria). CATENA 117, 60-72.

Sprafke, T., Obreht, I., 2016. Loess: Rock, sediment or soil – What is missing for its definition? Quat. Int., Kukla Loessfest 399, 198–207.

Stevens, T., Lu, H., Thomas, D.S.G. & Armitage, S.J., 2008. Optical dating of abrupt shifts in the Late Pleistocene East Asian monsoon. Geology 36, 415-418.

Stevens, T., Lu, H., 2009. Optically stimulated luminescence dating as a tool for calculating sedimentation rates in Chinese loess: comparisons with grain-size records. Sedimentology 56, 911–934.

Stevens, T., Marković, S.B., Zech, M., Hambach, U., Sümegi, P., 2011. Dust deposition and climate in the Carpathian Basin over an independently dated last glacial–interglacial cycle. Quat. Sci. Rev. 30, 662–681.

Sträßler, M., 1998. Klimadiagramme zur Köppenschen Klimaklassifikation. Klett-Perthes, Gotha.

Stuut, J.B., Smalley, I.J., O'Hara-Dhand, K., 2009. Aeolian dust in Europe: African sources and European deposits. Quaternary International 198, 234-245.

Sun, Y.B., Lu, H.Y., An, Z.S., 2006. Grain size of loess, palaeosol and Red Clay deposits on the Chinese Loess Plateau: significance for understanding pedogenic alteration and palaeomonsoon evolution. Palaeogeogr. Palaeoclimatol. Palaeoecol. 241, 129–138.

159

Svirčev Z., Marković S., Krstić S., Plavša J. (2009): Surface Freshwater Quality State in Vojvodina and Proposal for the WFD Monitoring System based on Some Biological Elements. Faculty of Natural Sciences, University of Novi Sad, Novi Sad.

Svirčev, Z., Marković, S.B., Stevens, T., Codd, G.A., Smalley, I., Simeunović, J., Obreht, I., Dulić, T., Pantelić, D., Hambach, U., 2013. Importance of biological loess crusts for loess formation in semi-arid environments. Quat. Int., Closing the gap - North Carpathian loess traverse in the Eurasian loess belt 6th Loess Seminar, Wrocław, Poland Dedicated to Prof. Henryk Maruszczak 296, 206–215.

Sweeney, M.R., Mason, J.A., 2013. Mechanisms of dust emission from Pleistocene loess deposits, Nebraska, USA. J. Geophys. Res. Earth Surf. 118, 1460-1471.

Thiel, C., Buylaert, J.-P., Murray, A., Terhorst, B., Hofer, I., Tsukamoto, S., Frechen, M., 2011. Luminescence dating of the Stratzing loess profile (Austria) – Testing the potential of an elevated temperature post-IR IRSL protocol. Quat. Int., Loess in Eurasia 234, 23–31.

Taylor, S.R., McLennan, S.M., 1985. The Continental Crust: its Composition and Evolution. Blackwell Scientific Publications Ltd., p. 312.

Thomsen, K.J., Murray, A.S., Jain, M., Bøtter-Jensen, L., 2008. Laboratory fading rates of various luminescence signals from feldspar-rich sediment extracts. Radiat. Meas. 43(9-10): 1474-1486.

Thompson, R., Oldfield, F.,1986. Environmental Magnetism. Allen& Unwin Ltd., London.

Timar, A., Vandenberghe, D., Panaiotu, E.C., Panaiotu, C.G., Necula, C., Cosma, C., others, 2010. Optical dating of Romanian loess using fine-grained quartz. Quat. Geochronol. 5, 143–148.

Timar-Gabor, A., Vandenberghe, D.A.G., Vasiliniuc, S., Panaoitu, C.E., Panaiotu, C.G., Dimofte, D., Cosma, C., 2011. Optical dating of Romanian loess: A comparison between silt-sized and sand-sized quartz. Quat. Int. 240, 62–70.

160

Timar-Gabor, A., Constantin, D., Marković, S.B., Jain, M., 2015. Extending the area of investigation of fine versus coarse quartz optical ages from the Lower Danube to the Carpathian Basin. Quat. Int. 388, 168-176.

Torrent, J., Schwertmann, U., Fechter, H., Alferez, F., 1984. Quantitative relationships between soil color and hematite content. Soil Sci. 136, 354–358.

Torrent, J., Liu, Q., Bloemendal, J., Barrón, V., 2007. Magnetic enhancement and iron oxides in the upper Luochuan loessepaleosol sequence, Chinese Loess Plateau. Soil Sci. Soc. Am. J. 71, 1570-1578.

Tošić, I., Hrnjak, I., Gavrilov, M.B., Unkašević, M., Marković, S.B., Lukić, T., 2014. Annual and seasonal variability of precipitation in Vojvodina Serbia. Theor. Appl. Climatol. 117, 331–341.

Trinkaus, E., Moldovan, O., Milota, Ş., Bîlgăr, A., Sarcina, L., Athreya, S., Bailey, S.E., Rodrigo, R., Mircea, G., Higham, T., Ramsey, C.B., Plicht, J. van der, 2003. An early modern human from the Peştera cu Oase, Romania. Proc. Natl. Acad. Sci. 100, 11231– 11236.

Tsoar, H., Pye, K., 1987. Dust transport and the question of desert loess formation. Sedimentology 34, 139-153.

Tzedakis, P.C., 1999. The last climatic cycle at Kopais, central Greece. J. Geol. Soc. 156, 425–434.

Tzedakis, P.C., Hooghiemstra, H., Pälike, H., 2006. The last 1.35 million years at Tenaghi Philippon: revised chronostratigraphy and long-term vegetation trends. Quat. Sci. Rev., Critical Quaternary Stratigraphy 25, 3416–3430.

Tzedakis, P.C., Lawson, I.T., Frogley, M.R., Hewitt, G.M., Preece, R.C., 2002. Buffered Tree Population Changes in a Quaternary Refugium: Evolutionary Implications. Science 297, 2044–2047.

Újvári, G., Kovács, J., Varga, Gy., Raucsik, B., Marković, S.B., 2010. Dust flux estimates for the Last Glacial Period in East Central Europe based on terrestrial records of loess deposits: a review. Quat. Sci. Rev. 29, 3157-3166.

161

Újvári, G., Kok, J.F., Varga, G., Kovács, J., 2016. The physics of wind-blown loess: Implications for grain size proxy interpretations in Quaternary paleoclimate studies. Earth-Sci. Rev. 154, 247–278.

Újvári, G., Stevens, T., Svensson, A., Klötzli, U.S., Manning, C., Németh, T., Kovács, J., Sweeney, M.R., Gocke, M., Wiesenberg, G.L.B., Markovic, S.B., Zech, M., 2015. Two possible source regions for Central Greenland last glacial dust. Geophys. Res. Lett. 2015GL066153.

Újvári, G., Varga, A., Balogh-Brunstad, Z., 2008. Origin, weathering, and geochemical composition of loess in southwestern Hungary. Quat. Res. 69, 421–437.

Újvári, G., Varga, A., Ramos, F.C., Kovács, J., Németh, T., Stevens, T., 2012. Evaluating the use of clay mineralogy, Sr–Nd isotopes and zircon U–Pb ages in tracking dust provenance: An example from loess of the Carpathian Basin. Chem. Geol. 304–305, 83–96.

Újvári, G., Varga, A., Raucsik, B., Kovács, J., 2014a. The Paks loess-paleosol sequence: A record of chemical weathering and provenance for the last 800ka in the mid- Carpathian Basin. Quat. Int. 319, 22–37.

Újvári, G., Molnár, M., Novothny, Á., Páll-Gergely, B., Kovács, J., Várhegyi, A., 2014b. AMS 14C and OSL/IRSL dating of the Dunaszekcső loess sequence (Hungary): chronology for 20 to 150 ka and implications for establishing reliable age–depth models for the last 40 ka. Quat. Sci. Rev. 106, 140–154.

Ünal-İmer, E., Shulmeister, J., Zhao, J.-X., Tonguç Uysal, I., Feng, Y.-X., Duc Nguyen, A., Yüce, G., 2015. An 80 kyr-long continuous speleothem record from Dim Cave, SW Turkey with paleoclimatic implications for the Eastern Mediterranean. Sci. Rep. 5, 13560.

Unkašević, M., Tošić, I., Obradović, M.H., 2007. Spectral analysis of the ‘‘Koshava’’ wind. Theor. Appl. Climatol. 89, 239–244.

Vandenberghe, J., Pissart, A., 1993. Permafrost changes in Europe during the last glacial. Permafrost Periglac. Process. 4, 121–135.

162

Vandenberghe, J., An, Z., Nugteren, G., Lu, H., Van Huissteden, J., 1997. A new absolute timescale for the Quaternary climate in the Chinese loess region based on grain size analysis. Geology 25, 35–38.

Vandenberghe, J., Huijzer, B., Mücher, H., Laan, W., 1998. Short climatic oscillations in a western European loess sequence (Kesselt, Belgium). J. Quat. Sci. 13, 471–485.

Vandenberghe, J., 2013. Grain size of fine-grained windblown sediment: A powerful proxy for process identification. Earth-Sci. Rev. 121, 18–30.

Vandenberghe, J., Markovič, S.B., Jovanovič, M., Hambach, U., 2014. Site-specific variability of loess and palaeosols (Ruma, Vojvodina, northern Serbia). Quat. Int. 334– 335, 86–93.

Varga, A., Újvári, G., Raucsik, B., 2011. Tectonic versus climatic control on the evolution of a loess–paleosol sequence at Beremend, Hungary: an integrated approach based on paleoecological, clay mineralogical, and geochemical data. Quat. Int. 240, 71–86.

Vasiliniuc, Ş., Vandenberghe, D.A.G., Timar-Gabor, A., Panaiotu, C., Cosma, C., van den Haute, P., 2012. Testing the potential of elevated temperature post-IR IRSL signals for dating Romanian loess. Quat. Geochronol. 10, 75-80.

Vasiliniuc, S., Vandenberghe, D.A.G., Timar-Gabor, A., Van Den haute, P., 2013. Conventional IRSL dating of Romanian loess using single aliquots of polymineral fine grains. Radiat. Meas. 48, 60-67.

Veres, D., Lane, C.S., Timar-Gabor, A., Hambach, U., Constantin, D., Szakács, A., Fülling, A., Onac, B.P., 2013. The Campanian Ignimbrite/Y5 tephra layer – A regional stratigraphic marker for Isotope Stage 3 deposits in the Lower Danube region, Romania. Quat. Int., Advancing Pleistocene and Holocene climate change research in the Carpathian-Balkan region 293, 22–33.

Virlet D'Aoust, P.T., 1858. Observations sur un terrain drigine météorique ou de transport aerien qui existe au Mexique et sur le phénomene des trombes de poussiere auquel il doit principalement son origine. Bulletin de la Société Géologique de France Deuxieme Série 15, 129-139.

163

Walker, M. J. C. (2005). Quaternary dating methods. John Wiley and Sons.

Wang, P., Tian, J., Lourens, L.J., 2010. Obscuring of long eccentricity cyclicity in Pleistocene oceanic carbon isotope records. Earth Planet. Sci. Lett. 290, 319–330.

Wang, X., Yang, Z., Chen, Y., Hu, J., Tang, L., Sheng, M., Løvlie, R., 2015. Deciphering magnetoclimatological patterns of late Early to early Middle Pleistocene loess– paleosol sequences in the western Chinese Loess Plateau. Glob. Planet. Change 130, 37–46.

Watanuki and Tsukamoto, 2001. A comparison of GLSL, IRSL and TL dating methods using loess deposits from Japan and China. Quat. Sci. Rev. 20, 847–851.

Wegwerth, A., Ganopolski, A., Ménot, G., Kaiser, J., Dellwig, O., Bard, E., Lamy, F., Arz, H.W., 2015a. Black Sea temperature response to glacial millennial-scale climate variability. Geophys. Res. Lett. 42, 2015GL065499.

Wegwerth, A., Ganopolski, A., Ménot, G., Kaiser, J., Dellwig, O., Bard, E., Lamy, F., Arz, H.W., 2015b. IRDc and accumulation rates of IRDc on sediment core M72/5_628-1.

Weidenbach, F., 1952. Gedanken zur Lößfrage. Eiszeitalter und Gegenwart 2, 25-36.

Whalley, W.B., Marshall, J.R., Smith, B.J., 1982. Origin of desert loess from some experimental observations. Nature 300, 433-435.

Wintle, A.G., 1973. Anomalous Fading of Thermoluminescence in Minerals. Nature 245(5421): 143-144.

World Meteorological Organization, 1996. Climatological Normals (CLINO) for the Period 1961–1990. Secretariat of the World Meteorological Organization, Geneva.

Worm, H.-U., 1998. On the superparamagnetic—stable single domain transition for magnetite, and frequency dependence of susceptibility. Geophys. J. Int. 133:201–206.

Wright, J.S., 2001a. “Desert” loess versus “glacial” loess: quartz silt formation, source areas and sediment pathways in the formation of loess deposits. Geomorphology 36, 231- 256.

164

Wright, J.S., 2001b. Making loess-sized quartz silt: data from laboratory simulations and implications for sediment transport pathways and the formation of ‘desert’ loess deposits associated with the Sahara. Quat. Int. 76-77, 7-19.

Wunsch, C., 2006. Abrupt climate change: An alternative view. Quat. Res. 65, 191–203.

Xiao, J., Porter, S., An, Z., Kumai, H., Yoshikawa, S., 1995. Grain size of quartz as an indicator of winter monsoon strength on the loess plateau of central China during the last 130,000 yr. Quat. Res. 43, 22–29.

Yaalon, D.H., Dan, J., 1974. Accumulation and distribution of loess-derived deposits in the semi-desert and desert fringe area of Israel. Zeitschrift für Geomorphologie Supplementband 20, 91-105.

Yang, S., Ding, Z., 2006. Winter-spring precipitation as the principal control on

predominance of C3 plants in Central Asia over the past 1.77 Myr: evidence from d13C of loess organic matter in Tajikistan. Palaeogeogr. Palaeoclimatol. Palaeoecol. 235, 330–339.

Yang, S.Y., Li, C.X., Yang, D.Y., Li, X.S., 2004. Chemical weathering of the loess deposits in the lower Changjiang valley, China, and paleoclimatic implications. Quat. Int. 117, 27–34.

Yang, S., Ding, F., Ding, Z., 2006. Pleistocene chemical weathering history of Asian arid and semi-arid regions recorded in loess deposits of China and Tajikistan. Geochim. Cosmochim. Acta 70, 1695–1709.

Yang, S., Ding, Z., 2014. A 249 kyr stack of eight loess grain size records from northern China documenting millennial-scale climate variability. Geochem. Geophys. Geosystems 15, 798–814.

Yang, S., Ding, Z., Li, Y., Wang, X., Jiang, W., Huang, X., 2015. Warming-induced northwestward migration of the East Asian monsoon rain belt from the Last Glacial Maximum to the mid-Holocene. Proc. Natl. Acad. Sci. 112, 13178–13183.

Yin, Q., Berger, A., 2015. Interglacial analogues of the Holocene and its natural near future. Quat. Sci. Rev. 120, 28–46.

165

Zaady, E.L.I., Offer, Z.Y., 2010. Biogenic soil crusts and soil depth: a long-term case study from the Central Negev desert highland. Sedimentology 57, 351-358.

Zachos, J., Pagani, M., Sloan, L., Thomas, E., Billups, K., 2001. Trends, Rhythms, and Aberrations in Global Climate 65 Ma to Present. Science 292, 686–693.

Zech, M., Buggle, B., Leiber, K., Marković , S.B., Glaser, B., Hambach, U., Huwe, B., Stevens, T., Sümegi, P., Wiesenberg, G., Zöller, L., 2009. Reconstructing quaternary vegetation history in the Carpathian Basin, SE Europe, using nalkane biomarkers as molecular fossils: problems and possible solutions, potential and limitations. Eiszeitalter und Gegenwart, Quaternary Science Journal 58, 148–155.

Zech, R., Zech, M., Marković, S., Hambach, U., Huang, Y., 2013. Humid glacials, arid interglacials? Critical thoughts on pedogenesis and paleoclimate based on multi-proxy analyses of the loess–paleosol sequence Crvenka, Northern Serbia. Palaeogeogr. Palaeoclimatol. Palaeoecol. 387, 165–175.

Zeeden, C., Hark, M., Hambach, U., Marković, S.B., Zöller, L., 2007. Depressions on the Titel loess plateau: form – pattern – genesis. Geogr. Pannon. 11, 4–8.

Zeeden, C., Hambach, U., Veres, D., Fitzsimmons, K., Obreht, I., Bösken, J., Lehmkuhl, F., in press. Millennial scale climate oscillations recorded in the Lower Danube loess over the last glacial period. Palaeogeogr. Palaeoclimatol. Palaeoecol.

Zeeden, C., Kels, H., Hambach, U., Schulte, P., Protze, J., Eckmeier, E., Marković, S.B., Klasen, N., Lehmkuhl, F., 2016. Three climatic cycles recorded in a loess-palaeosol sequence at Semlac (Romania) – Implications for dust accumulation in south-eastern Europe. Quat. Sci. Rev. 154, 130–142. doi:10.1016/j.quascirev.2016.11.002

Zeeden, C., Krauß, L., Kels, H., Lehmkuhl, F., 2017. Digital image analysis of outcropping sediments: Comparison to photospectrometric data from Quaternary loess deposits at Şanoviţa (Romania) and Achenheim (France). Quat. Int. 429, 100-107.

166

167

Supplementary Material Material Chapter 5

Supplementary Figure 5.1. The grain-size categories normalised to <63 µm related to the pedostratigraphy

168

Supplementary Table 5.1: Detailed major (wt%) and trace element (ppm) composition of the Orlovat section.

Depth (m) 8,15 8,05 7,95 7,85 7,75 7,65 7,55 7,45 7,35 7,25 7,15 7,05 6,95 6,85 6,75 6,7 6,6 Sample Elements no 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34

Na2O % 1,172 1,176 1,156 1,337 1,430 1,286 1,336 1,445 1,341 1,365 1,410 1,489 1,435 1,422 1,31826 1,342 1,226 MgO % 5,288 4,593 4,613 4,367 4,201 4,080 4,126 4,318 4,337 4,156 4,301 4,319 4,344 4,508 4,69905 4,910 4,574

Al2O3 % 13,471 14,440 14,368 14,780 14,614 15,076 14,960 14,780 14,858 14,912 14,928 14,831 14,780 14,839 14,90581 14,761 14,997 SiO2 % 55,666 57,459 56,786 57,897 56,015 56,913 57,127 56,840 56,077 56,544 56,334 56,083 55,730 56,039 55,70023 55,815 56,147 P2O5 % 0,213 0,243 0,242 0,221 0,246 0,266 0,253 0,268 0,252 0,260 0,236 0,240 0,230 0,226 0,21450 0,222 0,224 K2O % 2,491 2,599 2,581 2,634 2,728 2,717 2,703 2,659 2,647 2,556 2,543 2,488 2,482 2,466 2,44893 2,377 2,385 CaO % 15,318 12,652 13,396 11,862 13,663 12,558 12,330 12,596 13,318 13,089 13,103 13,433 13,842 13,294 13,51091 13,366 13,265

TiO2 % 0,952 1,010 1,018 1,016 1,019 1,005 1,017 1,017 1,028 1,036 1,015 1,027 1,031 1,079 1,07478 1,070 1,043 MnO % 0,104 0,109 0,108 0,107 0,110 0,109 0,113 0,105 0,108 0,107 0,107 0,108 0,108 0,113 0,11280 0,111 0,112 FeO % 5,325 5,720 5,733 5,779 5,974 5,989 6,036 5,972 6,034 5,975 6,023 5,980 6,018 6,012 6,01473 6,027 6,029 Rb ppm 74,6 80,25 79,65 82,85 82,8 85 84,7 83,85 84,05 83,2 83,55 83,3 82,9 83,95 83,55 84,55 85,15 Sr ppm 197,05 173,2 180,1 173,95 175,85 174 176,3 184,75 185,35 183,05 186,25 190,25 192,45 195,3 202,25 203,6 185,85 Ba ppm 307,5 326,5 333,5 355 331 353 359 353 361,5 356 360 355,5 362,5 349 362,5 369,5 341 Pb ppm 16 18,65 17,1 17,3 17 17,5 17,05 16,5 18,2 16,9 16,45 17,7 18,05 17,15 15,05 17,9 17,35 Th ppm 12,05 12,05 12,85 12,6 13 13 12,8 13,25 13,15 13,4 12,75 13,35 12,5 13,7 12,45 13,4 13,05 Zr ppm 313,5 343,7 329,4 343,9 354,5 359,4 376,95 381 375,95 409,55 390,1 394 401,45 408,7 404,75 438,3 400,9 Nb ppm 12,45 13,15 12,75 13,25 14,05 13,5 13,3 13,25 14,55 13,9 13,5 13,8 14,85 13,6 14,8 14,4 12,7 Y ppm 28,4 28,55 30,15 32 30,05 31,8 32,4 29,95 33,9 32,95 32,2 32,15 32,35 30,45 30,6 33 30,95 V ppm 87,9 96,2 88,9 104,7 99,95 103,8 97,75 109,95 101,35 106,55 105,1 105,65 106,4 106,5 98,25 101 118,3 Cr ppm 81,2 94 89,95 92,5 89,2 95 100,3 93,3 93,2 98,95 97,95 101,9 97,85 101,85 98,55 117,85 98,8 Ni ppm 35,45 39,2 38,95 42,65 42,2 44,4 43,5 42,1 43,1 42,7 43,2 42,95 41,1 41,55 43,1 42,75 43,35 Cu ppm 18,05 17,35 18,95 19,3 22,25 22,3 19,2 20 20,95 20,45 20,4 18,85 17,75 19,6 19,45 19,85 20,35 Zn ppm 63,3 68,9 66,9 69,75 69,25 71,2 72,35 70,95 70,95 71,3 72,3 71,45 71,65 71 71,4 72,15 71,35 Ga ppm 12,65 15,5 14,7 14,75 16,35 16,6 14,95 15,15 16,1 15,5 15,95 15,65 15,5 16,35 15,6 14,6 15,05

169

Depth (m) 6,4 6,3 6,2 6,1 6,0 5,9 5,8 5,7 5,6 5,5 5,4 5,3 5,2 5,1 5,0 4,9 4,8 Sample Elements no 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51

Na2O % 1,305 1,311 1,274 1,301 1,210 1,208 1,246 1,299 1,233 1,227 1,194 1,192 1,269 1,237 1,380 1,330 1,386 MgO % 4,798 5,705 6,305 5,597 5,269 5,468 5,430 5,155 5,010 5,149 4,973 5,255 5,592 5,851 5,339 5,361 5,137

Al2O3 % 14,804 14,387 14,639 15,030 15,203 14,869 15,072 15,082 14,983 14,882 14,808 14,484 14,468 14,199 14,424 14,594 14,859 SiO2 % 55,670 54,273 54,766 56,500 57,407 56,385 57,147 58,041 57,986 57,510 57,658 56,126 55,469 55,332 55,801 56,178 57,493 P2O5 % 0,214 0,209 0,194 0,193 0,192 0,203 0,187 0,209 0,200 0,191 0,198 0,191 0,194 0,198 0,222 0,213 0,211 K2O % 2,335 2,269 2,315 2,350 2,373 2,320 2,348 2,346 2,350 2,289 2,358 2,275 2,234 2,193 2,224 2,240 2,273 CaO % 13,736 14,856 13,514 11,913 11,113 12,515 11,543 10,888 11,153 11,848 11,882 13,610 14,003 14,264 13,809 13,200 11,777

TiO2 % 1,043 1,036 0,967 1,054 1,033 1,009 1,009 0,989 1,042 1,007 1,005 1,015 1,000 1,020 1,019 1,038 1,023 MnO % 0,109 0,102 0,109 0,116 0,120 0,117 0,115 0,114 0,112 0,109 0,108 0,107 0,106 0,105 0,104 0,107 0,107 FeO % 5,987 5,852 5,917 5,946 6,080 5,907 5,903 5,878 5,931 5,788 5,815 5,745 5,667 5,600 5,678 5,740 5,735 Rb ppm 83,4 83,05 85,95 87,85 87,95 82,65 83,65 81,6 80,1 78,55 78,75 76,3 74,4 72,1 73,6 75,25 77,9 Sr ppm 191,6 213,95 216,85 201,25 193,7 201 193,4 183,45 178,05 181,6 180,1 190,2 193,2 197,35 185,35 183,6 175,5 Ba ppm 323 337,5 342,5 357,5 381 365 339,5 353,5 348 344,5 359,5 336,5 353,5 336,5 317,5 332 336,5 Pb ppm 17,95 19,95 18,85 17,15 18,4 18,2 17,25 19,35 17,7 17,5 16,45 18,9 17,25 16,35 17,65 17,5 17,75 Th ppm 13,05 13,2 12,85 13,05 12,9 11,9 12,6 12,45 12,15 12,35 11,7 12,05 11,1 11,75 11,35 12,05 12,4 Zr ppm 403,3 391 373,6 374,05 371,05 379,8 360,1 369,2 367,95 389,55 367,7 369 370,3 376,95 350,9 382,15 378,15 Nb ppm 14,4 13,25 13,5 14 13,85 13,2 13 13,6 13,6 13,85 13,05 13,4 12,35 13,45 13,35 13,5 12,95 Y ppm 30,7 30,55 29,85 32 31,35 29,6 30,75 30,85 29,65 31,7 29,4 29,95 30,5 29,35 28,45 30,7 31,65 V ppm 95,85 97,15 105,3 107,25 107,3 97,25 105,2 104,6 103,2 99,8 96,4 97,9 84,55 96,45 86,65 98,3 92,25 Cr ppm 117,4 100,4 102,35 98,85 106,45 111,65 107,65 102,1 124,7 106,6 111,35 104,1 96,25 119,05 85,85 107,15 92,05 Ni ppm 42,7 43,55 45,4 47,9 49,7 47,55 48,2 48 45,95 44,4 44,25 42,9 42,15 41,2 41,65 42,8 44,05 Cu ppm 20,05 19,95 20,35 22,45 23 20,6 20,85 21,85 20,7 19,35 18,95 19,3 17,9 16,95 16,8 17,5 17,8 Zn ppm 71,35 69,1 71 72,3 73,55 71,7 74 72,05 70,55 70,3 70,2 68,3 66,9 65,1 64,8 66,85 69,25 Ga ppm 15,15 15,35 15,6 15,4 16,3 14,95 15,1 16,05 15,4 15,15 14,8 14,5 13,55 13,85 14,35 15,6 15,3

170

Depth (m) 4,7 4,6 4,5 4,4 4,3 4,1 4,0 3,9 3,8 3,7 3,6 3,5 3,4 3,3 3,2 3,1 3,0 Sample Elements no 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68

Na2O % 1,319 1,349 1,360 1,384 1,356 1,440 1,349 1,328 1,343 1,302 1,307 1,252 1,262 1,271 1,217 1,214 1,247 MgO % 5,364 5,118 5,362 5,873 5,078 4,505 4,243 4,235 4,236 4,271 4,132 4,052 4,003 4,003 3,915 3,943 3,843

Al2O3 % 14,664 14,605 14,600 14,551 14,841 15,134 15,422 15,340 15,049 15,210 15,143 15,021 15,295 15,207 15,238 15,007 14,755 SiO2 % 55,497 55,321 55,146 54,847 56,019 56,676 58,810 56,830 57,799 57,178 57,283 57,218 57,795 57,532 57,438 56,604 56,653 P2O5 % 0,216 0,216 0,211 0,226 0,238 0,236 0,235 0,240 0,248 0,241 0,257 0,249 0,251 0,264 0,262 0,272 0,270 K2O % 2,358 2,273 2,247 2,243 2,311 2,334 2,427 2,373 2,311 2,324 2,334 2,358 2,392 2,357 2,382 2,346 2,311 CaO % 13,564 14,180 14,139 13,803 13,016 12,485 10,274 12,461 11,960 12,351 12,458 12,682 11,868 12,233 12,388 13,438 13,882

TiO2 % 1,013 1,012 1,022 1,056 1,047 1,050 1,043 1,014 1,045 1,028 1,041 1,048 1,043 1,028 1,055 1,045 1,054 MnO % 0,107 0,108 0,108 0,118 0,120 0,121 0,124 0,123 0,119 0,122 0,121 0,121 0,125 0,127 0,128 0,134 0,128 FeO % 5,899 5,816 5,807 5,898 5,975 6,019 6,071 6,056 5,888 5,973 5,924 6,000 5,968 5,979 5,978 5,997 5,858 Rb ppm 77,15 76,25 76,6 78,75 80,1 82,5 83,25 80,5 77,7 78,3 78,2 77,95 79,95 80,35 79,3 78,75 76,5 Sr ppm 178,15 179,45 184,5 186,95 175,9 172,05 163,3 166,1 166,2 165,95 163,45 158,95 157,55 156 155,85 153,95 153,55 Ba ppm 361,5 343,5 325 348 333 352 382,5 365 358 351,5 346 363,5 374 348 337 326,5 345 Pb ppm 17,45 16,75 17,85 16,35 17,35 16,8 19,75 18,45 17,5 17,15 17,3 18,3 17,1 17,95 19 15,95 17,15 Th ppm 11,7 11,85 12,7 13,05 12,95 13,95 13,2 12,8 12,9 13,15 12,5 13,25 13,05 13,65 13 12,85 12,55 Zr ppm 339,6 363,85 378,2 441,5 404,55 414 409,55 398,05 402 419,75 413,7 417,8 402,6 410,2 417,1 437,3 430,85 Nb ppm 12,8 12,85 13,4 14,65 12,85 14,05 13,4 13,75 12,5 14 13,55 13,25 13,85 14,45 14,2 13,85 14,35 Y ppm 29,7 30 31,15 33,9 32,4 33,1 34,05 31,9 32,05 33,6 31,25 32,55 32,25 31,65 31,9 31,85 32,1 V ppm 91,45 94,3 95,7 99,2 108,95 105,25 85,8 105,9 101,4 108,7 103,95 93,35 98,7 99,2 105,4 103,2 102,3 Cr ppm 99,35 105,7 91,9 99,7 95,2 106,9 99,2 94,05 97,6 103,5 102,95 115,65 104,4 93,95 111,6 98,15 97,8 Ni ppm 42,5 41,05 41,6 45,45 44,35 46,15 46,95 47,3 45,8 46,7 46,3 45,2 46,2 45,6 46,2 46,85 43,7 Cu ppm 18,9 16,95 18,15 20,35 20,5 20,55 20,9 20,5 17,7 18,35 18,8 19,45 19,85 18,8 18 18,65 18,4 Zn ppm 68,5 66,85 69,6 71,3 73,35 73,1 73,7 73,4 69,3 72,4 72,2 71,2 71,35 71,75 70,15 70,3 67,8 Ga ppm 15,2 14,25 15,45 15,25 15,1 15,8 16,35 16,05 16,1 15,7 15,6 15,1 15,5 16,2 15,75 14,85 15,65

171

Depth (m) 2,9 2,8 2,7 2,6 2,5 2,4 2,3 2,2 2,1 2,0 1,9 1,8 1,7 1,6 1,5 1,4 1,3 Sample Elements no 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85

Na2O % 1,262 1,211 1,264 1,243 1,174 1,212 1,255 1,235 1,177 1,197 1,119 1,122 1,103 1,069 1,121 1,123 1,163 MgO % 3,848 3,713 3,755 3,745 3,731 3,759 3,810 3,802 3,771 3,754 3,842 3,724 3,840 3,841 3,866 3,866 3,927

Al2O3 % 14,643 14,314 14,434 14,460 14,215 14,187 13,882 13,693 14,042 14,085 14,169 14,189 14,146 14,246 13,782 13,546 13,542 SiO2 % 55,794 56,736 56,427 57,263 55,741 54,733 54,001 54,160 55,352 55,553 55,845 56,060 56,443 56,259 55,339 54,257 54,587 P2O5 % 0,294 0,283 0,289 0,279 0,303 0,276 0,264 0,268 0,276 0,274 0,268 0,281 0,274 0,276 0,271 0,274 0,273 K2O % 2,269 2,233 2,247 2,240 2,211 2,225 2,143 2,126 2,187 2,214 2,207 2,233 2,201 2,257 2,179 2,168 2,155 CaO % 14,816 14,658 14,620 13,828 15,653 16,691 17,769 18,043 16,376 16,204 15,787 15,671 15,455 15,343 16,866 18,260 17,834

TiO2 % 1,037 1,033 1,047 1,059 1,090 1,030 1,057 1,002 1,034 0,973 1,005 0,993 0,984 0,996 0,988 0,987 0,993 MnO % 0,133 0,128 0,131 0,132 0,126 0,125 0,123 0,119 0,121 0,121 0,123 0,119 0,120 0,123 0,118 0,112 0,115 FeO % 5,906 5,691 5,787 5,749 5,757 5,762 5,696 5,551 5,664 5,626 5,635 5,609 5,434 5,589 5,470 5,408 5,411 Rb ppm 77,3 74,75 74,65 75,2 73,45 72,9 72,3 71,7 73,2 75,1 75,8 77,3 75,8 76,5 74,3 72,25 72,95 Sr ppm 153,45 154,2 153,8 153,25 156,75 157,1 160,65 164,8 160,7 158,2 155,95 148,8 150,3 152,1 159,75 164,8 167,5 Ba ppm 355,5 326,5 326,5 328,5 307 328 318,5 352 315,5 323 308,5 324 330,5 330,5 312 313,5 301 Pb ppm 16,3 15,8 17,4 16,15 14,7 17,35 15,95 15,55 17,45 16,85 16,9 16,15 15,05 16,55 14,65 16,2 15,1 Th ppm 12,3 12,35 13,35 12,95 11,7 12,6 12,4 11,65 12,4 12,1 11,85 12,05 12,05 11,95 12,05 12,5 12 Zr ppm 431,35 478,8 479,45 462,3 485,85 414,3 452,55 396,8 437,7 417,85 418,3 368,35 395,85 387,2 408,5 375,2 398,75 Nb ppm 14,1 14,25 14,6 14,25 14,45 13,8 14 12,65 13,4 13,05 13,55 13,4 14,45 13,7 13 12,7 12,3 Y ppm 32,45 34,5 33 34,1 31,9 31,85 33,95 30,6 32 31,3 30,55 31,1 29,85 29,05 28,95 29,5 30,45 V ppm 103,7 100,95 98 90 89,35 97,55 92,35 91,4 92,4 82,65 99,45 96,35 91,6 89,75 83,45 89,65 87,8 Cr ppm 92,1 113,1 93,2 103,3 107,15 87,1 96,5 92,9 118,35 94,8 90,45 89,7 95,15 89,85 91,95 85,2 85,95 Ni ppm 45,85 45 45,55 43,6 44,1 43,25 44,3 41,2 42,9 42,15 41,8 40,45 40,85 39,35 39,3 39,15 37,75 Cu ppm 19,7 19,9 17,25 17,45 17,05 16,8 17,5 16,95 16,9 17,75 18,25 19,8 19,55 18,5 18,3 18,95 18,15 Zn ppm 69,65 66,3 67,95 66,65 67,15 65,5 64,4 62,9 65,7 66,35 66,35 66,45 65,6 65,2 64,25 63,35 64 Ga ppm 14,95 14,3 15,25 14,5 14,4 13,3 14,2 14,2 13,8 14,45 14,2 14,45 15,25 14,85 13,85 13,75 13,75

172

Depth (m) 1,2 1,1 1,0 0,9 0,7 0,6 0,5 0,4 Sample Elements no 86 87 88 89 90 91 92 93

Na2O % 1,070 1,069 1,080 1,110 1,088 1,121 1,126 1,180 MgO % 4,034 4,017 4,846 5,595 5,801 7,230 7,758 6,798

Al2O3 % 13,546 13,717 13,548 13,379 12,927 13,442 13,987 14,490 SiO2 % 54,023 54,949 53,400 52,853 51,502 53,259 55,027 56,125 P2O5 % 0,257 0,257 0,244 0,222 0,216 0,204 0,191 0,189 K2O % 2,162 2,176 2,143 2,116 1,998 2,065 2,143 2,210 CaO % 18,365 17,269 18,246 18,272 20,390 16,391 13,279 12,400

TiO2 % 0,975 0,965 0,963 0,969 0,898 0,934 0,973 0,987 MnO % 0,114 0,117 0,113 0,115 0,106 0,114 0,110 0,115 FeO % 5,454 5,463 5,417 5,369 5,075 5,239 5,406 5,508 Rb ppm 73,1 73,95 72,15 71,6 68,05 70,5 74,2 76,55 Sr ppm 170,3 161,05 172,2 182,55 204,75 215,25 199,65 190,7 Ba ppm 304,5 313 329 333 280 308,5 292,5 357 Pb ppm 15,35 16,3 15,15 16,2 12,85 16,45 14,6 16,2 Th ppm 11,4 11,65 10,8 10,8 10,3 10,9 10,8 10,9 Zr ppm 365,25 382,65 360 346,8 335,2 341,95 338,15 346,35 Nb ppm 11,35 12,4 11,75 12,55 10,7 11,4 12,15 13 Y ppm 28,55 31,05 27,4 27,5 26,85 27,55 27 28,35 V ppm 92,4 90,75 89,95 87,25 82,95 88,85 89,25 86,5 Cr ppm 82,25 87,05 79,85 84,5 76,55 83,8 82,15 85,6 Ni ppm 36,6 38,7 39,15 37,6 33,9 36,45 38,5 40,7 Cu ppm 18,7 17,05 18 18,55 16,15 17,85 17,65 18,35 Zn ppm 63,5 65,4 63,9 63,1 59,8 62,8 65,7 67,15 Ga ppm 12,95 12,55 13 13,55 12,2 12,95 13,6 13,7

173

Supplementary Material Chapter 6

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

Igor Obreht, Christian Zeeden, Ulrich Hambach, Daniel Veres, Slobodan B. Marković, Janina Bösken, Zorica Svirčev, Nikola Bačević, Milivoj B. Gavrilov, Frank Lehmkuhl

1. Sampling strategy

Two separate field campaigns were carried out in June and October, 2013. During the first campaign, a ca. 5.5 m long profile was sampled in the central part of the Stalać section (Supplementary Fig. 6.1). During the second campaign, four profiles on the south-west side of the brickyard situated at a slope position were sampled from the Stalać section (Supplementary Figs. 6.2 and 6.3). It was not possible to sample one continuous section because the brickyard has been mined actively, which had formed terraces. Special attention was paid that all of the sampled profiles were correlated by use of palaeosols and tephra layers; a certain amount of stratigraphic overlap was also included. Before sampling, 10-20 cm of exposed material was removed from all profiles, allowing for continuous, high- resolution, incremental sampling at all five sections. Samples were taken in 5 cm increments, with exception of the fourth profile, where 20 samples from the transition of the lowermost palaeosol to the loess layer above were sampled in 2.5 cm increments. In spring 2014, 13 samples of recent alluvium from the Zapadna Morava (4 samples), Južna Morava (5 samples) and Velika Morava (4 samples) Rivers were collected at 13 different locations (Supplementary Table 6.1).

174

Supplementary Table 6.1. Ni, Cr and magnetic susceptibility values from the Zapadna, Južna and Velika Morava rivers alluvium samples Magnetic susceptibility River Locality Cr [ppm] Ni [ppm] [10-8 m3/kg] Zapadna Morava Jasina 334.25 165.1 137 Zapadna Morava Šanac 396.6 272.6 200 Zapadna Morava Bosnjane 388.65 284.75 129 Zapadna Morava Maskare 438.4 288.65 165 Velika Morava Varvarin 241.45 138.95 78.1 Velika Morava Obrez 301.85 218.25 99.9 Velika Morava Trešnjevica 193.8 131.25 78.8 Velika Morava Confluence 211.05 115.75 105 Južna Morava Stalać 145.25 72.25 97.1 Južna Morava Branjina 135.25 39.5 88.6 Južna Morava Maletin 157.75 50.65 86.9 Južna Morava Donji Ljubeš 112.9 49.35 74.6 Južna Morava Confluence 125.1 52.5 106

Supplementary Figure 6.1. Photo of the stratigraphic section sampled (profile 1) during the first field campaign. The profile is marked with a red rectangle. It preserves a record of aaeolian sediment over the past ~350,000 years.

175

2. Stalać section – general features

The studied section (43°40'38.887"N, 21°25'4.18"E) is exposed in an active brickyard at the top of the Južna (South) Morava River terrace. The section is located about 3 km from the confluence of the Južna Morava and the Zapadna (West) Morava Rivers, forming the Velika (Great) Morava River (Fig. 6.1 in the original manuscript). For the period from 1981 to 2010, the mean annual temperature and mean annual precipitation of the nearby climate station at Kruševac were 11.4˚C (with high inter-annual variability) and 628 mm (with the main precipitation maximum during June, and a second maximum during November) (Republic Hydrometeorological Service of Serbia). The Stalać section is exposed in a wall that has a northeast to south-west orientation. The north-eastern and central part of the section show several loess and palaeosol layers, preserving proxy data for glacial and interglacial cycles. Six well developed, brown-reddish pedocomplexes, one hardly visible weak brown-grey palaeosol, and the modern soil, intercalated by loess layers, are exposed here. The south-western side of the exposure is developed on a slope and preserves the sediment formed from the last brown- red palaeosol to the modern soil. This younger sediment is preserved in much higher resolution. All palaeosols exposed at the south-west profile are inclined, following the slope exposure; they become thicker downslope. Palaeosols exposed on the slope exposure are bifurcating: the older brown-red palaeosol splits into two palaeosols and the younger (brown- grey palaeosol) splits into three. Supplementary Fig. 6.2 shows the correlation of the central part and south-western part of this section and Supplementary Fig. 6.3 shows the spatial correlation of the profiles. For this study, the upper part of the central profile (containing two brown-red palaeosols, a weak brown-grey palaeosol, the modern soil and the various intercalated loesses) and four profiles, spanning all the studied stratigraphic units from the south-western part.

176

Supplementary Figure 6.2. Photos of the stratigraphic sections, with labels to aid in correlation. Photos are of the central profile a) and the south-western part sampled during the second field campaign b) - profile 2-4 and c) – profile 5. Sampled areas of the profiles are marked with red rectangles. The tephra layer found in profiles 1, 2 and 3 corresponds to the same layer.

177

Supplementary Figure 6.3. Correlation of the profiles: a) profile 1 (as in Supplementary Fig. 6.2), b) profiles 2-4 (as in Supplementary Fig. 6.2), c) indication of profile 5. Due to semicircular excavation shape of this site, profile 5 is not shown here but on Supplementary Fig. 6.2.

3. Stratigraphy of the studied profiles

Profile 1 is 5.55 m long. Thicknesses and depths are reported from the base of the section upward, with zero being the bottom of the section. At the base, 0.4 m of loess (unit L4) is exposed. Above it is a strongly developed, brown-red Cambisol with strongly expressed vertic characteristics (S3); it has a total thickness of 1.25 m (from 0.4 to 1.65 m). The lowermost 0.4 m of the palaeosol has a brown-reddish B horizon. From 0.8 to 1.3 m, a strongly developed, brown-red B horizon is exposed and it is darker n colour than layer below. At the top of this layer, secondary carbonate features are developed. Both horizons have a blocky structure typical for vertic horizons. From 1.3 to 1.65 m, a red-brown Ah horizon with a granular structure is preserved. Above this palaeosol pedocomplex, from 1.65 to 3 m, is a layer of light yellow loess (L3). At 2.6 m, one krotovina was observed but not sampled. From 3 to 3.45 m, another Cambisol with 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). 178

Profile 2 presents the oldest (stratigraphically lowest) part of the profiles on the slope. At the bottom the first 0.2 m were built up by loess. A well-developed reddish palaeosol (0.2 m – 1 m) is formed above the S2LL2 loess layer and present the S2SS2 interstadial soil. On top of this soil 0.3 m of loess were deposited covering a weakly developed 0.15 m thick deep red palaeosol (S2SS1). Above 1.45 m, pale yellow loess (L2) is exposed showing some black organic spots (1.45 – 2.9 m). The grey and coarse tephra layer is situated on the top of profile (~2.9 – 2.95 m).

The bottom of profile 3 begins ca. 0.2 m below the tephra layer. The tephra itself is ca. 0.05 m thick. Above follows a thick loess layer (3.85 m) containing some carbonate concretions in the lower part (L2). The loess layer is intercalated by weak organic layers at ca. 3.05-3.15 m and 3.50-3.60 m. On top, a chernozem-like pedocomplex (S1) with calcareous root channels was formed (4.1 – 4.45).

Profile 4 starts in loess unit L2 (0 – 0.5 m) and continues to the ca. 0.85 m thick palaeosol S1. Above the palaeosol, a 1.30 m thick loess layer is exposed. From 2.65 m to the top of the sampled profile (3.05 m) a palaeosol is developed.

The bottom of profile 5 corresponds to the upper palaeosol from profile 4. Here, this palaeosol has a thickness of 1.00 m. It is followed by a darker loess layer (1.00 – 1.85 m). Above, another palaeosol horizon (1.85 – 2.65 m) is developed. A brown loess layer is exposed from 2.65 – 3.7 m. On top, the modern soil is exposed. The top of the modern soil is heavily influenced by human activities, thus only 30 cm of its lowermost part were sampled.

4. 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. 6.4 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-

179

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. 6.4).

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. 6.4).

More challenging is the splicing of profiles 4 and 5. Profile 4 did not reach the end of the upper L1SS2 palaeosol while profile 5 started just at the beginning of the same palaeosol unit with no overlap to the loess layer below. The lower limits of the palaeosols do not have a fit as good as between profiles 3 and 4. Thus, the splicing was not performed at the bottom of the 180

palaeosol but within the palaeosol where most of the proxies showed similar values (ca. 0.25 m below the end of profile 4 and 0.2 m above the beginning of profile 5; Fig 6.4). It may be possible that a few centimeters of the sediment are present twice or are missing, but generally data variation is low. Thus, we argue that this correlation does not suffer from major issues. The final composite profile is presented at Figs. 6.5 and 6.6.

Supplementary Figure 6.4. Splicing of profiles 1-5. Dashed black lines present the splicing points. Light lines of records present the part of profiles that were not used in the composite profile. Note that the upper part of profile 1 is compressed, revealing a different sedimentation rate. The difference in pattern between profiles 1 and 2 arises from different sedimentation rates; tephra occurrence provides a certain marker in L2 facilitating correlation between profiles 1 and 2.

181

Supplementary Figure 6.5. Grain-size distribution of each particle size class and cumulative distribution from all grain-size classes (each colour in the right box represent a corresponding class in µm) of the composite profile.

Supplementary Figure 6.6. Magnetic susceptibility, frequency dependent magnetic susceptibility, L*, a* and b* values related to pedostratigraphy of the composite profile.

182

5. Chronology and age model

Geochronology of loess-palaeosol sequences in the Middle Danube Basin has generally been established mainly by 14C and luminescence dating for the younger time periods (Fuchs et al., 2008; Hatté et al., 2013; Schmidt et al., 2010; Stevens et al., 2011). 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 (Basarin et al., 2014; Buggle et al., 2009; Fitzsimmons et al., 2012; Marković et al., 2012b; Necula et al., 2015; Necula and Panaiotu, 2008). 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 (Kostić and Protić, 2000). 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 proposed mutual climate connections between Central Balkans and Middle Danube Basin12 have to be reconsidered. In general, the stratigraphic framework of loess and the labeling of loess-palaeosol layers in the Middle Danube Basin is simple because the sequences were formed by relatively continuous deposition of aeolian dust (Marković et al., 2008, 2015) with only rare exceptions (Marković et al., 2014a; Obreht et al., 2015). The labeling of stratigraphic units follows the established scheme for the European loess-palaeosol sequences presented by Marković et al. (2015), using ‘L’ for loess units and ‘S’ for palaeosol units, while the ordinal numbers indicate the order of increasing age. Although the stratigraphic framework seems simple, the past climate characteristics of the central Balkans are not well understood. Thus, the determination of secure pedostratigraphic markers for simple stratigraphic linking of profiles is more complicated. For example, the correlation of visible layers is becoming challenging for the south-western section of the studied outcrop since the palaeosols there are splitting downslope.

183

One of approaches for obtaining age models used in Southeastern Europe is orbital tuning (Basarin et al., 2014; Marković et al., 2012b). Basarin et al. (2014) 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 (Marković et al., 2012b; Marković et al., 2015; Vandenberghe et al., 2014). 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 (especially the grain size trends (Supplementary Fig. 6.5), high Cl concentration (Supplementary Fig. 6.7) and microscopically observed volcanic glass shards (Supplementary Fig. 6.8)). The results from microprobe geochemical analyses of glass shards (Supplementary Fig. 6.9 and Supplementary Table 6.3) clearly relate this layer to the Campanian Ignimbrite/Y-5 tephra (~39 ka (Constantin et al., 2012; Fitzsimmons et al., 2013; Veres et al., 2013)) coming from the Campi Flegrei caldera in Italy, and identified so far in many locations throughout the Mediterranean Sea, in loess archives in the Lower Danube (Constantin et al., 2012; Fitzsimmons et al., 2013; Veres et al., 2013), and throughout records in the Balkans (Morley and Woodward, 2011) (Leicher et al., 2015). This volcanic ash layer has not been observed in the χ trend (albeit recorded in the Cl data), in contrast to the L2 tephra that is clearly visible in the data (Supplementary Fig. 6.6). The reason for this lies in the peralkaline trachytic composition of the Campanian Ignimbrite/Y-5 tephra. The χ of the Campanian Ignimbrite/Y-5 tephra does not exceed ~200*10-8 m3/kg (Fitzsimmons et al., 2013), thus having already lower values than the bulk loess sediment at Stalać. Since this tephra layer is not visible by naked eye, it is very likely it appears as a crypto tephra mixed up with some degree with aaeolian silt. Beside these two tephra layers, below the studied profile and thus beyond the frame of this study, another potential tephra layer was observed in the

184

middle of the L4 unit. This tephra may correspond to the Bag tephra observed in the Middle Danube Basin (Marković et al., 2012b; Marković et al., 2015). This provides additional (albeit tentative) evidence that supports the established chronology.

Supplementary Fig. 6.7 presents the composite profile with tephra proxies (Cl and χ) correlated to the age model. The glass-shard chemical composition assigns the upper volcanic ash bed to the Campanian Ignimbrite/Y-5 tephra, with an age of ~39 ka (Constantin et al., 2012; Fitzsimmons et al., 2013; Veres et al., 2013), while the occurrence of the L2 tephra is tentatively related to the MIS 6. The age model is based on a simple correlation of palaeosols to odd MIS (Lisiecki and Raymo, 2005) (S0 – MIS 1; L1SS1 and L1SS2 – MIS 3, S1 – MIS 5, S2 – MIS 7 and S3 – MIS 9) and loess layers to even MIS (L1LL1 – MIS 2, L1LL3 – MIS 4, L2 – MIS 6, L3 – MIS 8 and upper part of L4 – late MIS 10), as established for the Middle Danube Basin (Basarin et al., 2014; Buggle et al., 2009; Marković et al., 2012b; Marković et al., 2015). An exception is layer L1SS1LLL1 that is attributed (at least in its lower part) to a mixture of Campanian Ignimbrite/Y-5 ash and coarse dust particles (Supplementary Fig. 6.7), and it is used as an additional tie point (Supplementary Table 6.2).

To test the age-model additional evidence (albeit preliminary) from luminescence dating (Marković et al., 2006) is also briefly discussed. For equivalent doses (De) determination, fine-grained (4-11µm) polymineral samples were measured in a Risø TL/OSL DA 20 reader at the Cologne Luminescence Lab. The post infrared infrared stimulated luminescence (pIRIR) protocol by Thiel et al. ( 2011) and the central age model (Galbraith et al., 1999) were used. Prior to IR stimulation temperature tests and dose recovery tests were conducted satisfactorily. Dose rates were determined through the measurement of radionuclide concentrations in a high-purity germanium gamma-ray spectrometer, corrected by conversion and attenuation factors of Guerin et al. ( 2011), Brennan 2003), Mejdahl (1979), and the measured water content. A potassium content of 12.5±0.5% was assumed (Huntley and Lamothe, 2001). Alpha efficiencies of 0.13±0.01 for C-L3780 and 0.11±0.01 for C-L3784 and C-L3786 were employed. The preliminary luminescence ages are presented in Supplementary Fig. 6.7 and Supplementary Table 6.5 (the reader is referred to Bösken et al. (2017) for further details) and all ages are in very good agreement with the proposed age model. Nevertheless, the preliminary luminescence ages are in low resolution, and thus cannot be employed at this stage in constraining the millennial-scale climate oscillations, but they incontestably indicate validity of the age-model discussed. 185

Supplementary Figure 6.7. Correlation of height scale and the resulting age assignment of different units at Stalać. Bulk sediment geochemical proxy Cl and magnetic susceptibility respectively, provided useful proxies for the identification of tephra layers. On the left, preliminary luminescence ages (Marković et al., 2006) are presented.

Supplementary Table 6.2. Tuning points of Stalać age model to Marine Isotope Stages based on LR04 stack (Lisiecki and Raymo, 2005)

Tuning point Height (m) Age (ka) Beggining of profile 0 350 L4/S3 transition 0.4 337 S3/L3 transition 1.65 300 L3/S2 transition 3 243 S2/L1 transition 3.45 191 L1/S1 transition 8.9 130 S1/L1LL3 transition 9.75 71 L1LL3/L1SS2 transition 11.05 57 L1SS2/L1LL2 transition 12.15 40 L1LL2/L1SS1 transition 12.85 38 L1SS1/L1LL1 transition 13.7 29 L1LL1/S0 transition 14.65 11 End of profile 15 8

186

Supplementary Table 6.3. Major oxide geochemical results from microprobe analyses of glass shards from upper crypto tephra layer (the Campanian Ignimbrite/Y-5) at the Stalać section. Data are presented as raw values. Analytical settings used for determining the glass-shard major oxides composition is presented at the Supplementary Table 6.4. - Location SiO2 TiO2 Al2O3 FeO MnO MgO CaO Na2O K2O P2O5 Cl Total

Stalać 58.79 0.35 18.09 2.77 0.22 0.35 1.60 5.63 6.56 0.05 0.65 95.06 59.31 0.40 18.10 2.88 0.24 0.31 1.69 5.75 7.00 0.03 0.71 96.42 60.12 0.43 18.45 2.81 0.23 0.33 1.67 6.14 6.73 0.01 0.72 97.64 58.08 0.33 17.50 3.25 0.08 0.66 2.35 2.88 8.88 0.19 0.30 94.50 60.12 0.42 18.48 2.83 0.23 0.30 1.62 6.35 6.81 0.04 0.73 97.92 59.98 0.42 18.47 2.86 0.23 0.30 1.64 6.49 6.69 0.01 0.69 97.77 59.33 0.42 18.06 2.84 0.21 0.32 1.66 6.17 6.53 0.05 0.76 96.35 60.59 0.41 18.61 2.95 0.21 0.35 1.68 6.46 6.87 0.04 0.74 98.91 59.35 0.42 18.24 2.79 0.22 0.33 1.61 6.08 6.88 0.05 0.70 96.67 59.47 0.39 18.25 2.79 0.19 0.31 1.64 6.30 6.52 0.04 0.74 96.64 58.70 0.41 17.93 2.80 0.25 0.35 1.59 6.28 6.43 0.05 0.74 95.52 58.68 0.39 17.99 2.77 0.20 0.33 1.74 5.85 6.93 0.07 0.73 95.67 58.13 0.43 18.08 2.79 0.22 0.32 1.77 5.55 7.00 0.05 0.76 95.10 59.47 0.41 18.20 2.87 0.21 0.32 1.57 6.29 6.66 0.03 0.67 96.70

187

58.87 0.38 17.86 3.33 0.11 0.71 2.46 2.96 9.28 0.18 0.28 96.43 58.60 0.44 18.30 2.71 0.21 0.30 1.63 5.84 6.53 0.06 0.74 95.35 60.65 0.46 18.53 3.04 0.23 0.31 1.60 6.51 6.91 0.01 0.68 98.94 58.96 0.42 18.59 2.77 0.22 0.30 1.70 5.84 6.60 0.04 0.75 96.18 58.41 0.38 17.88 2.84 0.19 0.29 1.61 5.74 6.60 0.05 0.72 94.70 58.74 0.38 18.20 2.83 0.21 0.32 1.76 5.89 6.97 0.04 0.75 96.10 59.58 0.39 18.18 2.83 0.25 0.31 1.69 5.99 6.95 0.03 0.78 96.99 60.72 0.42 18.49 2.94 0.22 0.32 1.64 6.28 7.24 0.00 0.74 99.02 59.87 0.39 18.50 3.01 0.22 0.34 1.76 6.47 6.85 0.03 0.78 98.21 58.45 0.36 17.38 3.05 0.09 0.68 2.29 3.29 9.12 0.10 0.32 95.13 59.91 0.36 17.85 3.14 0.12 0.65 2.34 3.27 9.13 0.14 0.31 97.23 59.24 0.39 18.47 2.90 0.25 0.33 1.69 5.52 7.04 0.02 0.67 96.52 58.27 0.37 17.90 3.16 0.09 0.72 2.45 3.40 8.59 0.09 0.30 95.34 58.78 0.44 17.96 3.29 0.12 0.72 2.51 3.37 9.21 0.13 0.30 96.83 58.40 0.35 17.82 3.18 0.11 0.71 2.36 2.91 9.03 0.13 0.31 95.32 58.18 0.36 17.71 3.33 0.04 0.72 2.47 2.92 9.63 0.17 0.27 95.81 57.76 0.42 17.69 2.86 0.24 0.33 1.72 5.86 6.52 0.02 0.78 94.21 58.52 0.44 18.32 2.80 0.21 0.34 1.63 5.64 6.74 0.07 0.64 95.34

188

60.23 0.46 18.37 2.96 0.20 0.37 1.66 6.19 6.91 0.08 0.74 98.17 60.01 0.47 18.23 2.90 0.20 0.33 1.65 6.45 6.71 0.01 0.78 97.75 58.48 0.44 18.23 2.78 0.28 0.33 1.72 5.97 6.81 0.03 0.69 95.77 60.42 0.42 18.30 2.81 0.20 0.31 1.57 6.20 6.83 0.09 0.68 97.83 60.80 0.39 18.38 2.92 0.22 0.35 1.64 6.28 6.96 0.02 0.67 98.63 59.79 0.39 18.38 2.78 0.21 0.32 1.59 6.25 6.77 0.05 0.68 97.21 58.77 0.46 17.75 2.78 0.22 0.32 1.74 4.99 6.83 0.13 0.72 94.71 59.54 0.39 18.27 2.87 0.23 0.35 1.69 6.44 6.55 0.03 0.77 97.13 58.13 0.38 17.94 3.30 0.08 0.73 2.55 2.94 9.33 0.18 0.30 95.87 59.61 0.43 18.02 2.92 0.27 0.34 1.67 6.35 7.24 0.05 0.75 97.65 60.24 0.40 18.30 2.98 0.21 0.33 1.69 6.32 6.96 0.04 0.75 98.23 58.94 0.32 18.08 3.32 0.13 0.73 2.51 3.03 9.47 0.19 0.30 97.02

Average 59.25 0.40 18.14 2.94 0.19 0.41 1.84 5.39 7.34 0.07 0.63 96.60 St. Dev. 0.83 0.04 0.29 0.18 0.06 0.16 0.33 1.30 1.03 0.05 0.18 1.30

189

Supplementary Table 6.4. Analytical settings used for determining the glass-shard major oxides composition at Bayerisches GeoInstitut, Bayreuth University. Order of measuring elements (first to last): Na, Si, K, Ca, Fe, Mg, Al, P, Ti, Mn, Cl-.

Element Standard Measuring time

(Standard-Block) (peak/bkgr.)

Si /Kα C-Forsterite 30/15

(Cameca Block5)

Mn /Kα C-MnTiO3 40/20

Ti /Kα (Cameca Oxides (2)) 40/20

Al /Kα C-Spinel 30/15

(Cameca Block5)

Fe /Kα Fe-GP40 30/15

(GP40)

Mg /Kα C-Enstatite 30/15

(Cameca Block5)

Ca /Kα C-Wollastonite 30/15

(Cameca Oxides (2))

K /Kα C-Orthoclase 30/15

(Cameca Oxides (2))

Na /Kα C-Albite 10/5

(Cameca Oxides (2))

P /Kα C-Apatite 60/30

(Cameca Oxides (2))

Cl- /Kα C-Vanadinite 30/15

(Cameca Oxides (2))

190

Supplementary Figure 6.8. Microscopical photos of volcanic glass shards from cryptotephra L1SS1LLL1 layer. The images were taken with magnification factor 50. The sediment was suspended in water and viewed in transmitted light.

191

Supplementary Figure 6.9. Total alkali - Silica diagram (modified after Anechitei-Deacu et al. (2014)) showing the geochemical correlation between the Campanian Ignimbrite tephra identified at Stalać and other regional occurrences, including proximal pyroclastic flow (Civetta et al., 1997) and plinian fall deposits in Italy (Signorelli et al., 1999), as well distal fine ash occurrences within Mediterranean marine records (Pyle et al., 2006) and in the Russian loess (Pyle et al., 2006), the terrestrial sequence at Caciulatesti in southern Romania (Veres et al., 2013), and Urluia (Fitzsimmons et al., 2013) and Rasova – Valea cu Pietre (Anechitei-Deacu et al., 2014) loess/palaeosol sequences in the Lower Danube loess. Inset: scanning electron microscope (SEM) image of the Campanian Ignimbrite glass shards from Stalać.

192

Supplementary Table 6.5. Summary of the De, dose rate and resulting pIRIR age data (Bösken et al., 2017). Water content was obtained in the laboratory. Equivalent doses (De) are shown as result of central age model (Galbraith et al., 1999). Standard errors are indicated. Ages are expressed with a 1-sigma error range.

Sample Lab code Grain-size Water Total dose rate De (Gy) Age name (µm) content (%) (Gy/ka) CAM/COM (ka) St 3 C-L3780 4-11 11.8 ± 5.9 4.07 ± 0.2 648±35 135±7 St 7 C-L3784 4-11 8.6 ± 4.3 4.06 ± 0.2 621±32 135±6 St 9 C-L3786 4-11 8.0 ± 4.0 3.71 ± 0.2 245±13 58±2.4 St 10 C-L3787 4-11 10.7 ± 5.4 3.97 ± 0.2 192±10 40.8±2 St 11 C-L3788 4-11 7.8 ± 3.9 4.26 ± 0.2 149±8 29.5±1.2

6. Principle of geochemical tracing palaeo-river dranaige network over the Central Balkan

The grain-size records generally show remarkably coarser grain distributions (Supplementary Fig. 6.5) compared to the other sections in Southeastern Europe (Antoine et al., 2009; Bokhorst et al., 2011; Marković et al., 2008; Obreht et al., 2015; Vandenberghe et al., 2014) covering the same time interval (Fig. 6.3 in the original manuscript). This indicates that the Stalać section reflects mainly an enhanced dust contribution from proximal sources. The density distribution curve indicates primary aeoilan deposition (Supplementary Fig. 6.10). The possible source areas for the Stalać section are the valleys of Južna Morava, Zapadna and Velika Morava Rivers (Fig. 6.1 in the original manuscript). These rivers present the main drainage network of the Central Balkan, and Stalać is located at the confluence of these rivers. The geology in the catchment of these rivers is characterised by metamorphic and igneous rocks providing minerals with χ about one order of magnitude higher than in loess deposits along the Danube and in China (Basarin et al., 2011). At the Stalać section, χ is not closely following the pedostratigraphy (Supplementary Fig. 6.6), as in most other sections in the Middle and Lower Danube Basins (Buggle et al., 2009, 2014; Fitzsimmons et al., 2012; Marković et al., 2015). Supplementary Table 6.1 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 (Evans et al., 2003; Wang et al., 2015)

193

may play a role, though an unclear relation of χ and other palaeoenvironmental proxies (specifically the Ni and Cr content, and the L* and a*) do not support a dominant wind or provenance component. The χ signal here is probably resulting from available source material, wind intensity and also pedogenic processes including carbonate presence/absence. The Zapadna Morava alluvium has higher χ values compared to the Južna Morava (Supplementary Table 6.1), and an increase in particle contribution from the Zapadna Morava would be reflected as an increase in χ, and vice versa. The middle part of the Zapadna Morava River catchment is built from ultramafic rocks, which have a high contribution of Ni and Cr (Kabata-Pendias, 2010) and contain a high amount of magnetic iron minerals. Because Ni and Cr are considered more reliable proxies for provenance here than the χ, geochemical proxy data is used to discuss provenance. Elevated concentrations of Ni and Cr are found in the Zapadna Morava alluvium downstream from the ultramafic rock band (Supplementary Table 6.1). Consistence of Ni, Cr and χ changes within the section present a clear hint at periodic shifts in the domination of particles originating from the alluvium carried by the Zapadna and Južna Morava. Generally, lower values of Ni and Cr are demonstrating the predomination of the Južna Morava River, while higher values of Ni and Cr indicate domination of Zapadna Morava River particles.

Generally, changes in aeolian sediment provenance area are related to changes in wind directions, wind vigor and/or source area (Bokhorst et al., 2011; Obreht et al., 2015). However, the main source area is in the proximity of the Stalać section and changes in the wind direction will not change the main source area greatly. Since the provenance changes inferred from the trends in Ni and Cr represent the interaction between domination of river discharge from the Južna (lower Ni and Cr values) and Zapadna Morava (higher Ni and Cr), it may be used to trace the palaeo-river drainage system over the Central Balkan. Due to the different characteristics of its catchments, understanding of palaeo-river drainage system can be used to infer a signal of past precipitation change on a wider region (Supplementary Fig. 6.11). Under current conditions, the Zapadna and Južna Morava River catchments are affected by different precipitation regimes (Republic Hydrometeorological Service of Serbia). River discharge of the Zapadna Morava River is strongly influenced by the tributary Ibar River, which contributes with almost the same amount of water runoff as the Zapadna Morava. Their catchment areas receive annual precipitation above 800 mm, with some areas >1000 mm, especially in the higher mountains (Republic Hydrometeorological Service of Serbia). Thus, a

194

decrease in the rainfall amount in the western sectors of the Balkan Peninsula will have a significant effect on those rivers discharge rate. Even a gentle decrease in precipitation has a considerable effect on the discharge since it will affect both the Ibar and Zapadna Morava rivers and their corresponding catchments. Contrary, the Južna Morava River catchment area is located in the rain shadow of high mountains related to the Zapadna Morava River catchment, and it is mainly characterised by annual rainfall amounts of 500-600 mm, with some parts of enhanced annual precipitation not exceeding 800 mm (Republic Hydrometeorological Service of Serbia). Since this catchment area is composed of only one major river as a main drainage course, decrease in already low precipitation will have much weaker influence. According to that, the discharge of the Južna Morava River would be less sensitive to precipitation changes over its catchment, while the discharge of the West Morava River would strongly depend on the precipitation regime over its catchment area. Therefore, higher values of Ni and Cr at the Stalać section would indicate more precipitation over the Zapadna Morava catchment relative to the Južna Morava catchment. Also, higher precipitation would enhance erosion processes over the Zapadna Morava catchment compared to the Južna Morava catchment, and therefore additionally enhance the contribution of particles from its catchment area. Contrary, low values of Ni and Cr would indicate mostly similar distribution of the precipitation over these two catchment areas. Generally, it can be assumed that domination of Južna Morava River sediments (high Ni and Cr values) during glacials would indicate very low precipitation over both catchment areas. That would point to very dry conditions where the precipitations did not reach into the interior of Balkans. However, domination of particle from Južna Morava River over interglacials would indicate equally high precipitation over these two catchment areas. That means that air masses that bring precipitation crossed even over the high Dinaric Mountains west from the Južna Morava catchment.

7. Inferred past climate and environmental changes from the Central Balkans

Palaeosol S3 (MIS 9) is the strongest developed palaeosol at the Stalać profile for the studied time interval. High clay contributions, redness and vertic characteristics indicate strong weathering with hematite formation during dry and hot summers (Torrent et al., 1984). This indicates a strong influence of Mediterranean climate. Further to the south, referring to pollen- based palaeoenvironment reconstruction from Lake Ohrid (Sadori et al., 2016), this period is 195

characterised by the expansion of mesophilous vegetation, indicating warm and dry conditions for interglacial standards over the whole Balkans.

In general, the grain-size analysis is a well-accepted method in inferring palaeoclimate forcing upon the formation of loess-palaeosol sequences (Hao et al., 2012; Vandenberghe, 2013), 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. 6.5). High values of Ni and Cr (Supplementary Fig. 6.11) point towards dry conditions in south and central Serbia and possibly more moisture over mountainous areas to the west, as reflected by an enhanced particle contribution from the Zapadna Morava River catchment. Decrease in the percentage of coarse particles upwards points to warmer conditions, while gradual decrease in Ni and Cr values suggests slight increase in humidity over south and central Serbia during late MIS 8. Contrary to a benthic δ18O isotope stack (Lisiecki and Raymo, 2005), our data suggest that the coldest and driest interval was the middle part of MIS 8 (rather than late MIS 8). However, this is in agreement with the MEDSTACK planktic 18O data (Wang et al., 2010), as well as the Lake Ohrid pollen data (Sadori et al., 2016).

The subsequent palaeosol S2 denotes another interglacial period that corresponds to MIS 7 and is much less developed compared to the S3. It still shows typical features of a palaeosol formed under warm and dry summer conditions, but it is less abundant in clay and with weaker redness values. Likely, Mediterranean climate influence still played an important role as determining factor for soil development during that time interval, but its influence is decreasing in comparison to MIS 9. Data from the Ohrid lacustrine record (Sadori et al., 2016) shows a very high interglacial variability during MIS 7, with three expansions of trees interrupted by two intervals with herb expansion. It is more challenging to interpret palaeo- river drainage network and it’s dynamics during interglacials. Higher contribution of the Južna Morava alluvium particles observed during all interglacials (Supplementary Fig. 6.11) likely indicates similar amounts of precipitation over Central Balkan, showing that althoug decerasing with time, Meditteranean influence was stronger than today during all three past interglacials. Alternatively, the closer Južna Morava valley might be favoured as a source area

196

simply due to weaker wind dynamic during interglacials. However, first scenario seems to be more plausible becouse of the vicinity of all river valleys suggested as souce areas.

The unit corresponding to the MIS 6 glacial period in the Stalać record is the L2. The abundance of sand in this unit implies strong wind dynamics and apparently the most severe conditions during past ~350,000 year. The geochemical tracing suggest that the L2 loess is mostly formed by particles from the Južna Morava River valley. This suggests very dry conditions over the whole Balkans. The Zapadna Morava River discharge was strongly reduced, whereas discharge from the Južna Morava catchment was not dramatically limited by the general aridization. According to grain-size distributions (Supplementary Fig. 6.5), the middle part of L2 was characterised by very dry and cold conditions, but at the onset and end of the glaciation environmental conditions may have been less severe. However, colour data (especially b* values; Supplementary Fig. 6.6) suggest different environmental conditions during the onset and the end of MIS 6. This is in agreement with the Lake Ohrid pollen data (Sadori et al., 2016) which suggest a change from grassland (189–161 ka) to steppe dominated environment (161–126 ka).

The palaeosol representing the MIS 5 (S1) shows remarkably different features compared to the older palaeosols at the Stalać section. In the Middle Danube Basin palaeosols linked to the last two interglacials are represented by chernozem type soil with no remarkable difference between S2 and S1 (Marković et al., 2012b). The Stalać section shows clear change in environmental conditions from a vertic layer in S2 to a chernozem-like S1. Similarities in genesis of S1 (MIS 5), L1SS1SSS2 and L1SS1SSS1 (MIS 3) pedocomplexes indicate broadly similar conditions during the last interglacial and MIS 3 interstadials at the Stalać section. High fine particle abundance (Supplementray Figs. 6.5 and 6.12) and high L* values (Supplementary Figs. 6.6 and 6.12) during the last glacial indicate relatively mild conditions compared to the previous glaciation. The only proxy indicating enhanced weathering during the formation of S1 compared to MIS 3 is the frequency dependent magnetic susceptibility. However, frequency dependent magnetic susceptibility in general shows a narrow range of variations compared to other East European sections (Buggle et al., 2014) and because of their different origin and properties compared to the Balkan loess, should be considered with caution. Contrary to the Stalać section, where MIS 3 palaeosol formations started well before 40 ka, palaeosols related to MIS 3 in the Middle Danube Basin are much weaker than interglacial S1 palaeosols, with the main soil formation phase after ~40 ka (Stevens et al., 197

2011) and only very weak soil development before (Fuchs et al., 2008; Schmidt et al., 2010; Stevens et al., 2011). An exception is the L1SS1LLL1 layer, which shows the coarsest grain- size distribution (Supplementary Fig. 6.5), remarkably higher Cl values (Supplementary Fig. 6.7) and presence of volcanic glass shards (Supplementary Fig. 6.8). 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 (Fitzsimmons et al., 2013; Veres et al., 2013) (Supplementary Fig. 6.9). This layer most probably represents a short accumulation event of volcanic ash and aaeolian silt, indicating strong adverse conditions during local impact of the Campanian Ignimbrite - Y5 tephra layer and possibly Heinrich event 4. However, more uniform and mild conditions during the late last glacial in Stalać are in agreement with results from another Central Balkans loess- palaeosol sequences section at Belotinac (Obreht et al., 2014).

198

Supplementary Figure 6.10. Density distribution curves of grain size analyses of representative samples from all stratigraphic units at each sampled profile. Representative samples of one unit from each profile are presented by colour and codes that correspond to the numbers from 1 to 35. Explanation of each sample code is presented in Supplementary Table 6.6 where it is shown to which layer, MIS, and height on single and composite profile (if presented) the samples correspond.

199

Supplementary Table 6.6. Codes related to the samples presented in Supplementary Fig. 6.10. The first column contains the code, the second presents the layer and MIS to which the samples correspond, the third presents the height of the samples on the original profile and the fourth column presents the height of the samples on the composite profile (if presented)

Code number Layer / MIS Height on profile [m] Height on composite profile [m] 1 L4 / MIS 10 0.35-0.40 0.35-0.40 2 S3 / MIS 9 1.15-1.20 1.15-1.20 3 L3 / MIS 8 1.85-1.90 1.85-1.90 4 S2 / MIS 7 3.15-3.20 3.15-3.20 5 L2 / MIS 6 4.00-4.05 - 6 S1 - L1SS1 / MIS 5-3 4.50-4.55 - 7 S0 / MIS 1 5.40-5.45 - 8 S2 / MIS 7 0.20-0.25 - 9 S2 / MIS 7 0.70-0.75 - 10 S2LL1 / MIS 7 1.05-1.10 - 11 S2 / MIS 7 1.35-1.40 - 12 L2 / MIS 6 1.90-1.95 3.85-3.90 13 L2 / MIS 6 2.60-2.65 4.55-4.60 14 L2 tephra /MIS 6 2.90-2.95 4.85-4.90 15 L2 tephra / MIS 6 0.20-0.25 4.90-4.95 16 L2 / MIS 6 0.50-0.55 5.20-5.25 17 L2 / MIS 6 1.20-1.25 5.90-5.95 18 L2 / MIS 6 2.00-2.05 6.70-6.75 19 L2 / MIS 6 3.05-3.10 7.75-7.80 20 L2 / MIS 6 3.60-3.65 8-30-8.35 21 S1 / MIS 5 4.20-4.25 - 22 L2 / MIS 6 0.10-0.15 - 23 S1 / MIS 5 0.55-0.60 8.95-9.00 24 S1 / MIS 5 1.125-1.15 9.525-9.55 25 L1LL3 / MIS 4 1.375-1.40 9.775-9.80 26 L1LL3 / MIS 4 1.75-1.80 10.15-10.20 27 L1LL3 / MIS 4 2.05-2.10 10.45-10.50 28 L1SS2 / MIS 3 2.85-2.90 - 29 L1SS2 / MIS 3 0.35-0.40 11.35-11.40 30 L1SS2 / MIS 3 0.90-0.95 11.90-11-95 31 L1LL2 tephra / MIS 3 1.20-1.25 12.20-12.25 32 L1SS1 / MIS 3 1.85-1.90 12.85-12.90 33 L1SS1 / MIS 3 2.40-2.45 13.45-13.50 34 L1LL1 / MIS 2 3.10-3.15 14.10-14.15 35 S0 / MIS 1 3.85-3.90 14.85-14.90

200

Supplementary Figure 6.11. Ni and Cr concentrations presented on age scale. Green dashed lines present mean value of samples from the Great Morava alluvium sediment, red dashed lines present mean value of samples from the Zapadna Morava alluvium sediment and yellow dashed lines present mean value of samples from the Južna Morava alluvium sediment. Values located in the light red rectangle are under relative domination of the Zapadna Morava alluvium as source area, while values in light yellow rectangle are under relative domination of the Južna Morava alluvium as source area.

201

Supplementary Figure 6.12. Mean grain-size (µm) and L* values from the Stalać section compared to the normalised modelled Greenland ice-sheet volume (de Boer et al., 2014a). Note that the transition from MIS 7 to MIS 6 at the Stalać section (transition from S2 to L2 layer) has to be considered with caution as explained in Supplementary chapter 4.

202

Supplementary Material Chapter 7

Shift of large-scale atmospheric systems over Europe during late MIS 3 and its implications to modern human dispersal

Igor Obreht, Ulrich Hambach, Daniel Veres, Christian Zeeden, Janina Bösken, Thomas Stevens, Slobodan B. Marković, Nicole Klasen, Dominik Brill, Christoph Burow, Frank Lehmkuhl

1. Sampling strategy

The samples from the Urluia section were collected during a field campaign in spring 2014; the Vlasca section was sampled in autumn 2015. Profiles were freshly cleaned to allow for continuous, high-resolution, incremental sampling. Samples for sedimentology and environmental magnetism were taken in 2 cm increments. At the Urluia section, samples from 2.04 m to 8.50 m depth were taken as oriented samples for future palaeomagnetic studies. Oriented samples were not taken in the upper 2.04 m because strong bioturbation occurs ranging from mm to dm scale. Generally, during sampling large scale bioturbation structures (e.g. krotovinas) were avoided. Moreover, seven luminescence samples were taken at Urluia by hammering and extracting 15 cm long metal tubes into the sediment wall, and by collecting the surrounding material in a radius of 30 cm for dose rate determination.

2. Stratigraphy of studied profiles 2.1.Urluia The Urluia loess section (44.09417°N, 27.9031°E, 125 m a.s.l) is located on the southern Dobrogea plateau in southeastern Romania. In this area, up to few tens of m thick loess- palaeosol sequence covers Cretaceous limestone uplifted in Early to Mid Quaternary (Fitzsimmons and Hambach, 2014). The outcrop face at Urluia is ca. 800 m long and strikes NNW-SSE. This study focuses on the upper part of the loess sequence exposed in an abandoned quarry. The top of the section is formed by a well-developed ~1 m thick dark greyish-brown steppe soil followed by a diffuse transition towards loess from ~1.00-1.20 m, characterised by a transition towards lighter sediment and a decline of crumbling soil structure. From 1.20-1.70 m, yellow to yellow-brownish loess with a homogenous structure is exposed. This layer is

203

rich in pseudo-mycelia (small sub-mm/dm-scale) carbonate precipitation following mm-sized biogalleries) and krotovinas (cm to sub-dm sized burrows of steppe mammals) whose frequencies decreases with depth. Several loess dolls (pedogenetic carbonate concretions) are observed between 1.45-1.65 m, albeit they are notably present until 2.50 m. From 1.70-2.40 m, the loess becomes less brownish and more yellowish with depth, while below 2.40 m it becomes yellowish. The upper 2.50 m of the profile are generally strongly bioturbated on a cm to sub-dm scale. At ~3.55 m, a clear change in color is observed from yellowish to brown- yellowish. From 3.50-3.70 m loess is notably coarser in gain-size and darker than above. The loess below is generally homogenous until 5.15 m, and below this layer until 5.45 m loess is again brownish and the amount of root channels with humic in fillings increases. A diffuse ochre-colored horizon occurs between 5.45-5.65 m, followed bygreyish-yellow loess with brownish mm-sized spots spanning the section until 7.40 m depth. From 5.70-5.95 m a dm- scale horizontal krotovina filled with grey loess and surface soil material (not sampled) crosses the sampled wall. Along the loess wall at Urluia, in some parts the tephra layer is preserved in up to one meter thickness on palaeoslopes and a palaeodepression, and it can be followed along most of the wall. This tephra is related to the Campanian Ignimbrite tephra layer (Fitzsimmons and Hambach, 2014; Fitzsimmons et al., 2013). The sampled profile was chosen to be at a position which is least influenced by slope processes. Therefore, on the sampled wall the tephra is not very clearly standing out. However, based on detailed lithological characteristics and inter-profile correlation along the outcrop walls, at a depth between 6.15-6.2 m a crypto-tephra layer is suspected. From 7.40-7.50 m loess is pale reddish-brown, while at ~8.10 m a diffuse and bioturbated but clear border towards grey underlying loess occurs. 2.2.Vlasca The Vlasca section (44.39951° N, 27.86409° E, 37 m a.s.l.) is located in the Lower Danube Basin (Southeastern Romania), and was sampled in a loess cliff at the Danube between the villages Vlasca and Stelnica. The cliffs are 1.8 km long and are striking from SW to NE. The studied profile is in total 10.72 m high. Also at Vlasca, the top of the section is formed by a well-developed 0.8 m thick generally dark greyish-brown steppe soil; a 0.09 m rather thin middle brown horizon at the top is followed by a crumbled to subpolyedric textured middle brown horizon until ~0.30 m. Sediment from ~0.30 m until ~0.50 m has the same features as above, but it is lighter in color (light grey brown). From 0.50-0.68 m, the crumbled structure decreases. Down to 0.68 m 204

depth the sediment is strongly rooted. Below 0.68 m and until 0.80 m the rooting intensity decreases and a gradual change in color towards more brownish shades is observed. From 0.80-1.00 m, a fine powdery and lighter greyish-whitish loess layer is followed by a 0.12 m horizon with initial forming of loess dolls and decreasing frequency of whitish carbonate spots complete the base of the modern soil formation. From 1.12-1.39 m depth downward alight brown greyish powdery loess with subpolyedric crumbly matrix occurs. Ochre (to ochre-greyish in the lower part) typical loess follows below until 2.43 m depth, with abundant small krotovinas and a bigger one (ca. 10 cm in diameter) at ~1.60 m. From 2.43-3.40 m typical homogenous loess is exposed. This interval is followed by loess with a gentle increase in more crumbled structure until 7.70 m. From 7.70 m to 9.00 m the sediment is dark-ochre to brown in color and mm-sized dark organic spots are observed. The upper boundary to a tephra layer is very diffuse, but the pure tephra layer is exposed from 9.00-9.05 m. Below the tephra, the loess is generally sandier, and from 9.05-9.25 m it has a yellowish-brown-grey color, while it becomes greyish and more cemented below. Brownish loess appears from 9.80 to 9.90 m. From 9.90 until the end of the profile loess is again greyish. From 10.10 m downwards, the sand content and the degree of cementation rapidly increases.

3. Luminescence dating

3.1.Equivalent dose (De) measurements

Samples for De determination were prepared under subdued red illumination. When opening the steel cylinders, the first ~2 cm of sediment from both ends were removed. After drying the samples for 24 h at 50°C the gravimetric water content was calculated after weighing the samples and hence the loss of water. The dried samples were then treated with 10% hydrochloric acid, 10% hydrogen peroxide and 0.01 N sodium oxalate to remove carbonates, organic matter, and to disperse sediment aggregates. The 4-11 µm grain size fraction was isolated using Stokes’ law and by removing the clay fraction (< 4 µm) via centrifuging.

Continuous wave optically stimulated luminescence measurements (CW-OSL) were carried out on a Risø TL/OSL DA 20 reader. The reader is equipped with a 90Sr/90Y β-source and infrared (IR) LEDs emitting at 870 nm (FWHM = 40 nm). A 410 nm interference filter was used for signal detection. The pIR50IR290 SAR protocol was used (Buylaert et al., 2012; Thiel et al., 2011). The signal was integrated using the first 2.4 s of the stimulation curve minus a background derived from the last 25.6 s. A prior IR stimulation temperature test 205

(Buylaert et al., 2012) was performed on sample C-L3713. Furthermore, several dose recovery tests (DRTs) with the favourable prior IR stimulation temperature were performed using bleached samples (24 h solar simulator). Finally, a mean equivalent dose was determined using the central age model (CAM) and the common age model (COM) (Galbraith et al., 1999). Residual doses were assessed after bleaching the aliquots for 24 hours in a Hönle Sol2 solar simulator. Fading tests were conducted on sample C-L3711. For these, a laboratory beta dose of 186 Gy was used and the normalised luminescence signals were measured after different storage times between 100 and 800 minutes.

3.2.Dose rate measurements

Samples for dose rate determination were oven dried at 50 °C for at least 48 h, homogenised and packed into plastic cylinders for radionuclide measurements. Radionuclide concentrations were measured on a high-purity germanium gamma-ray spectrometer after a resting period of four weeks to compensate for radon emanation during pretreatment. The dose rates and ages were calculated using DRAC, v.1.2 (Durcan et al., 2015). Conversion factors of Liritzis et al. (2013) and an estimated water content of 10±5% were included. The internal β dose rate contribution was calculated by assuming a K content of 12.5±0.5% (Huntley and Baril, 1997) and a Rb content of 400±100 ppm (Huntley and Hancock, 2001). Attenuation factors of Bell (1980) and Guerin et al. ( 2012) were used. The cosmic dose rate was calculated following Prescott and Hutton 1994) considering the geographical position, altitude, sample depth and density of the overlying sediments. The α-efficiency was determined for six aliquots of two samples. Aliquots were thermally annealed (480°C) before administering α-doses up to 200 Gy in a Freiberg Instruments Lexsyg Research device. The laboratory dose of the samples was determined using β-irradiation in a Risø reader (pIRIR protocol). Finally, the α-efficiency was calculated as the ratio of the measured β-dose divided by the given α-dose (Kreutzer et al., 2014). A linear function was used for fitting. A mean a-value was used for all samples.

206

Figure S7.1. Dose recovery test (top) and prior IR stimulation temperature test (bottom). All tested samples are within the desired recovered/given dose ratio of 1.0±0.1. A plateau for prior IR stimulation temperatures is present between 50 and 170°C.

207

Figure S7.2. Consistency between luminescence ages (red squares and 1-σ uncertainty), correlative age model (grey line; black dots represent tie points) and the final age model (black line) of the Urluia section. Orange lines represent the timing and the depth of the Campanian Ignimbrite tephra layer.

208

Table S7.1. Summary of luminescence data for samples from Urluia (URL1). IRSL measurements were done on polymineralic fine grains (4-11

µm) on 9.8 mm aliquots using the pIRIR290 protocol (Buylaert et al., 2012; Thiel et al., 2011). An internal K content of 12.5 ± 0.5 % (Huntley and Baril, 1997) and Rb content of 400 ± 100 ppm (Huntley and Hancock, 2001) was assumed. The cosmic dose rates were calculated according to Prescott & Hutton (Prescott and Hutton, 1994), taking into consideration the altitude (125 m), latitude (44.094167°N) and longitude (27.9031°E) of the sampling site, as well as the density (1.7 g cm-3) and thickness of the overlying sediments. Ages are presented with 1-σ uncertainty.

b a d e Sample Depth n U (ppm) Th (ppm) K (%) Wmeasured Wused Ḋcosmic Ḋtotal DRT Residual De (Gy) OD (%) Age (ka) (m) (wt. %)c (wt. (Gy ka-1) (Gy ka-1) ratio dose %)c (Gy) C-L3716 1.50 10/10 2.4±0.13 8.83±0.54 1.3±0.01 - 10±5 0.18±0.02 4.02±0.22 - - 84.27±4.29 0.0±0.0 21.0±1.6 C-L3715 2.00 24/24 2.86±0.14 9.41±0.49 1.45±0.03 14.4 10±5 0.14±0.01 4.48±0.24 0.91±0.01 7.1±0.28 111.93±6.05 9.36±1.55 25.0±1.9 C-L3713 3.30 10/10 3.0±0.14 11.24±0.52 1.54±0.03 11.0 10±5 0.12±0.01 4.73±0.26 0.99±0.04 6.98±0.1 135.9±7.01 0.59±6 28.7±2.2 C-L3712 4.50 10/10 3.22±0.15 11.35±0.57 1.67±0.03 8.3 10±5 0.12±0.01 5.11±0.28 0.96±0.02 6.36±0.44 166.27±8.72 2.52±2.21 32.5±2.5 C-L3711 5.20 20/20 3.43±0.16 11.75±0.59 1.64±0.03 9.0 10±5 0.11±0.01 5.26±0.29 1.02±0.00 6.64±0.29 190.25±9.84 4.15±1.32 36.2±2.8 C-L3709 6.00 10/10 3.24±0.15 10.94±0.55 1.62±0.03 9.6 10±5 0.11±0.01 5.01±0.28 1.07±0.02 6.18±0.37 221.01±11.67 3.1±2.07 44.1±3.4 C-L3708 6.90 11/11 3.0±0.14 10.85±0.54 1.51±0.03 - 10±5 0.1±0.01 4.76±0.26 - - 258.24±13.3 1.05±3.41 54.2±4.1 a a-value measured: 0.136±0.02 b Number of accepted in relation to measured aliquots. c Gravimetric water content. d DRT = Dose Recovery Test. e OD = overdisperion, calculated using the Central Age Model (Galbraith et al., 1999b).

209

4. Chronology and age model 4.1.Tephra chronology

At both sections a tephra layer was observed. At Urluia, this layer is only preserved as the crypto-tephra in the sampled main profile, albeit it can be clearly observed and followed over a major part of section wall using lithostratigraphic markers. Overall, the tephra layer is thicker in depressions and on slopes (up to one meter in some parts of the Urluia section (Fitzsimmons et al., 2013)) of the primary relief at the time of deposition. Based on microprobe analyses and luminescence dating, previous investigations have related this layer to the Campanian Ignimbrite/Y-5 tephra (Fitzsimmons et al., 2013; Fitzsimmons and Hambach, 2014; Veres et al., 2013).

Microprobe analyses of the tephra layer at Vlasca were performed on samples that were taken ~700 m northeast along the Danube bank from the studied section. These investigations were performed before the sedimentological sampling of the Vlasca section discussed here, in order to establish a chronology of the Vlasca loess site. Figure S7.3 shows the scanning electron microscope images of the glass shards, other magmatic and detrital aeolian grains, and Table S2 shows the geochemical composition of tephra glass shards. According to the geochemistry of the glass shards, this layer is unambiguously related to the Campanian Ignimbrite tephra. Consequently, chronologies of both sections are grounded on the Campanian Ignimbrite/Y-5 tephra, a reliable chronological layer dated to 39.93 ± 0.1 ka BP (Scaillet et al., 2013).

210

Figure S7.3. Scanning electron microscope (SEM) images of the Campanian Ignimbrite glass shards from the Vlasca site.

211

Table S7.2. Major oxide geochemical results from microprobe analyses of glass shards of the tephra layer (Campanian Ignimbrite/Y-5) at the Vlasca section. Data are presented as raw values. Analytical settings used for determining the glass-shard major oxides composition is presented in the Table S7.3.

Sample SiO2 TiO2 Al2O3 FeO MnO MgO CaO Na2O K2O P2O5 Cl- Total STE1 61.07 0.38 18.17 2.88 0.09 0.55 2.17 4.06 8.04 0.061 0.346 97.81 STE2 60.43 0.40 18.54 2.75 0.18 0.33 1.68 6.3 6.67 0.021 0.729 98.02 STE3 61.01 0.44 18.65 2.84 0.23 0.40 1.62 6.5 7.47 0.010 0.741 99.91 STE4 61.23 0.33 18.11 2.68 0.12 0.47 1.98 4.41 8.3 0.103 0.409 98.13 STE5 61.55 0.37 18.23 2.92 0.11 0.58 2.22 4.54 7.98 0.133 0.367 99.00 STE6 60.49 0.45 18.4 2.76 0.25 0.36 1.67 6.03 7.88 0.024 0.722 99.03 STE7 60.82 0.37 18.67 2.84 0.19 0.32 1.73 6.35 7.43 0.039 0.718 99.48 STE8 60.53 0.43 18.58 2.92 0.20 0.34 1.7 6.27 7.34 0.059 0.711 99.08 STE9 60.77 0.45 18.57 2.8 0.26 0.31 1.64 6.55 7.17 0.044 0.785 99.35 STE10 61.37 0.45 18.8 2.85 0.21 0.34 1.65 6.53 6.8 0.033 0.692 99.73 STE11 61.3 0.44 18.61 2.82 0.25 0.35 1.62 6.69 7.28 0.047 0.748 100.16 STE12 61.19 0.48 18.71 2.86 0.29 0.29 1.62 6.77 6.85 0.025 0.708 99.79 STE13 60.84 0.44 18.48 2.89 0.22 0.35 1.67 5.9 8.12 0.054 0.708 99.66 STE14 60.86 0.43 18.66 2.93 0.22 0.36 1.72 6.65 6.84 0.108 0.726 99.51 STE15 61.18 0.43 18.7 2.91 0.24 0.33 1.79 5.91 7.67 0.016 0.710 99.88 STE16 60.77 0.37 18.73 2.73 0.23 0.31 1.64 6.61 6.75 0.056 0.739 98.93 STE17 60.42 0.43 18.44 2.89 0.18 0.34 1.62 6.56 6.85 0.066 0.726 98.53 STE18 60.73 0.35 18.2 3.4 0.13 0.71 2.52 3.49 9.2 0.151 0.328 99.21 STE19 60.91 0.40 18.56 2.84 0.17 0.35 1.73 6.28 7.12 0.051 0.727 99.15 STE20 61.25 0.41 18.58 2.87 0.21 0.33 1.7 6.69 7.09 0.046 0.727 99.90 STE21 60.54 0.45 18.57 2.84 0.23 0.37 1.65 6.52 6.96 0.042 0.740 98.92 STE22 59.22 0.38 18 3.45 0.11 0.79 2.67 3.08 9.46 0.110 0.291 97.56 STE23 59.84 0.38 18.55 2.83 0.24 0.32 1.59 6.65 6.71 0.044 0.743 97.89 STE24 62.09 0.33 18.47 2.69 0.13 0.48 1.94 5.03 7.92 0.078 0.457 99.62 STE25 60.6 0.43 18.39 2.88 0.23 0.35 1.58 6.31 6.69 0.034 0.739 98.24 212

STE26 61.43 0.36 18.45 3.04 0.11 0.58 2.18 5.12 7.55 0.068 0.339 99.22 STE27 62.22 0.38 18.54 2.94 0.13 0.59 2.16 3.43 7.19 0.110 0.346 98.04 STE28 60.36 0.39 18.68 2.8 0.23 0.35 1.66 6.52 6.9 0.069 0.720 98.67 STE29 60.73 0.42 18.55 2.81 0.20 0.31 1.62 6.61 6.95 0.051 0.733 98.98 STE30 61.05 0.40 18.75 2.89 0.24 0.35 1.6 6.8 7.35 0.029 0.742 100.19 STE31 59.84 0.38 18.08 2.62 0.21 0.37 1.66 5.58 7.19 0.093 0.459 96.47 STE32 60.4 0.42 18.46 2.83 0.22 0.32 1.62 6.2 6.82 0.014 0.738 98.05 STE33 60.46 0.45 18.45 2.89 0.18 0.29 1.71 6.63 7.02 0.069 0.695 98.84 STE34 60.45 0.43 18.39 2.79 0.21 0.35 1.8 6.39 6.96 0.091 0.685 98.55 STE35 61.03 0.39 18.31 2.92 0.23 0.36 1.65 6.51 7.09 0.054 0.735 99.28 STE36 60.58 0.45 18.35 2.91 0.23 0.30 1.72 6.66 6.87 0.046 0.703 98.82 STE37 61.71 0.38 18.53 2.69 0.17 0.39 1.66 5.92 7.53 0.017 0.527 99.53 STE38 61.67 0.38 18.41 2.75 0.15 0.51 2.06 4.53 8.27 0.076 0.378 99.18 STE39 60.79 0.44 18.61 2.8 0.24 0.36 1.72 6.38 7.32 0.017 0.711 99.39 STE40 60.86 0.41 18.44 2.81 0.26 0.30 1.75 6.67 7.25 0.008 0.697 99.46 STE41 61.2 0.39 18.7 2.84 0.21 0.34 1.92 5.85 7.14 0.005 0.751 99.35 STE43 60.9 0.41 18.62 2.8 0.23 0.32 1.68 6.22 7.43 0.064 0.696 99.37 STE44 60.31 0.42 18.44 2.86 0.21 0.29 1.61 6.6 6.8 0.059 0.728 98.33 Average 60.86 0.41 18.49 2.86 0.20 0.39 1.79 5.94 7.35 0.06 0.64 98.98 St. Dev. 0.57 0.04 0.19 0.15 0.05 0.11 0.25 1.00 0.63 0.03 0.15 0.77

213

Table S7.3. Analytical settings used for determining the glass-shard major oxides composition at Bayerisches GeoInstitut, Bayreuth University. Order of measuring elements (first to last): Na, Si, K, Ca, Fe, Mg, Al, P, Ti, Mn, Cl-.

Element Standard Measuring time (Standard-Block) (peak/bkgr.)

Si /Kα C-Forsterite 30/15 (Cameca Block5)

Mn /Kα C-MnTiO3 40/20 Ti /Kα (Cameca Oxides (2)) 40/20

Al /Kα C-Spinel 30/15 (Cameca Block5)

Fe /Kα Fe-GP40 30/15 (GP40)

Mg /Kα C-Enstatite 30/15 (Cameca Block5)

Ca /Kα C-Wollastonite 30/15 (Cameca Oxides (2))

K /Kα C-Orthoclase 30/15 (Cameca Oxides (2))

Na /Kα C-Albite 10/5 (Cameca Oxides (2))

P /Kα C-Apatite 60/30 (Cameca Oxides (2))

Cl- /Kα C-Vanadinite 30/15 (Cameca Oxides (2))

214

4.2. Age model

Obtaining a reliable age model can be achieved in several ways. In loess research, the most common ones are by numerical dating (using luminescence (Marković et al., 2014; Stevens et al., 2011; Timar et al., 2010) or radio carbon dating (Hatté et al., 2013; Újvári et al., 2014)) or by correlation to isotopic Greenland ice sheet δ18O or speleothem records (Yang and Ding, 2014; Zeeden et al., in press). Establishing an age mode relying on luminescence dating has serious limitations in precision. Constructing such the age model is challenging, especially when using ages with 2-σ uncertainty. Since climate proxy parameters of the studied sections show a pacing comparable to millennial scale oscillation, using a correlative age model can significantly reduce the uncertainties. Such short scale millennial oscillations are reported in Southeastern European lacustrine records (Feurdean et al., 2014; Panagiotopoulos et al., 2014), including the Black Sea (Shumilovskikh et al., 2014; Wegwerth et al., 2015a) (albeit Ménot and Bard (Ménot and Bard, 2012) reported a lack in D/O interstadial signatures in the mean annual lake surface temperatures), and also in speleothem records representing Mediterranean climate (Fleitmann et al., 2009; Ünal-İmer et al., 2015). Accordingly, recording millennial scale climatic oscillations in loess may be expected (Zeeden et al., in press). However, beside the correlative method, it is still not possible to relate fluctuations of loess proxies to certain interstadial-stadial events due to low resolution and lack in precision of dating techniques. Up to now, it has never been dated, and therefore convincingly demonstrated, that abrupt climate changes recorded in loess match individual D/O events. This problem is not only restricted to loess, but prevalent in many records (Wunsch, 2006). Since the abrupt changes recorded in loess and D/O events might not coeval, using a correlative age model might increase bias to non-climatic forcing (instead of increasing precision). In particular the grain-size can be influenced by many factors beyond a straightforward relation to climate changes. The grain-size distribution strongly depends on the proximity to source area, sediment availability, environmental conditions at the source, during transport and deposition, trapping mechanisms, and even on the basic physical background of dust particle transport and deposition mechanisms (Újvári et al., 2016). Moreover, it has been demonstrated that grain-size records from adjacent sections (only few tens of meters apart) at the Chinese Loess Plateau show different short term fluctuation, even if the long term patterns overlap (Stevens and Lu, 2009). Zeeden et al. (in press) proposed 215

frequency dependence of magnetic susceptibility (χfd) as a sensitive record for changes in sediment humidity and potentially the best proxy for indicating variations in D/O events scaling. However, here we show that albeit the general trends of the Urluia, Vlasca and Rasova (Zeeden et al., in press) sections are quite similar (Fig. S7.6), there are also notable differences in the amplitude of fluctuations and short term patterns. This indicates an important influence of local conditions to the extent of χfd oscillation, and therefore again might record a bias to changes not directly related to climate change in the wider region.

To obtain the age model in our study, we did not fully rely on the correlative age model or luminescence dating. Since we cannot certainly claim that D/O events are coeval with fluctuations in loess records, correlating our records to D/O events might increase a bias rather than precision of the age model. We compiled an age model that reduces the tie points to a minimum, and uses only the most certain tie points. In this way, the resolution of the age model decreases, but also the potential bias to error is minimised. We use the Campanian Ignimbrite tephra as a robust tie point at both sections, and only the correlations that are strongly supported by luminescence dating are used as tie points. Since the Vlasca section was not directly dated so far, this section is correlated to the general trends of proxy parameters at the Urluia section. We feel certain about this approach because both sections indicate very similar patterns in grain-size and environmental magnetic data. Details on the age models from both sections are presented in following chapters.

216

4.2.1. Age model based on luminescence dating

In order to obtain an age-depth model for the loess sequence at Urluia, seven luminescence samples are used in a first step (Fig. S7.2). The thorough investigation of the luminescence samples shows no problematic behavior. This demonstrates the general suitability of the pIRIR290 protocol for these samples. The pIRIR290 signal is argued to be not affected by fading (Buylaert et al., 2012; Thiel et al., 2011), which was supported by our fading measurements resulting in a negative mean g-value. Since the measured fading rates are likely to be a laboratory artifact (Buylaert et al., 2011), equivalent doses were not corrected for fading. The resulting ages increase constantly with depth. The layer of the Campanian Ignimbrite tephra dated to 39.93 ± 0.1 ka (Fitzsimmons et al., 2013; Lowe et al., 2012; Veres et al., 2013) presents a robust tie point and it is used as an independent test for the precision of the luminescence ages. The luminescence sample C-L3709 at 6.0 m, ~10 cm above the tephra layer, indicates an age of 44.1 ± 3.4 ka. Using the lower boundary of 1 σ uncertainty (Fig. S7.2), the time period shows slightly older age for the sediment deposited after the Campanian Ignimbrite tephra accumulation. The age is ~2-3 ka older than expected, and even if this does not indicate a perfect match, the ages might be assumed as reliable indications of the sedimentation timeframe on a more general scale, especially when considering 2-σ uncertainty (under which the ages overlap).

4.2.2. Correlative age model

It is challenging to compile the age model using luminescence ages with 2 σ uncertainty, since the uncertainty is considerably high (up to ~16.2 ka years for the C-L3708 sample and 2-σ). To reduce the uncertainty we compile a correlative age model based on the correlation of proxy data from Urluia to millennial scale palaeoclimatic patterns observed in δ18O data from the Greenland ice sheet (Andersen et al., 2004) (Fig. S7.4, Table S7.4). For the correlation to the Greenland δ18O record we used the frequency dependent magnetic susceptibility (χfd) data as suggested by Zeeden et al. (in press), but also paid attention to the patterns in grain-size distribution. Luminescence ages then cross-checked this model (Fig. S7.2). Obtaining an age-depth model for Vlasca (Fig. S7.5 and Table S7.4) was more challenging since the numerical time control of the luminescence dating is not given. However, the timing of the deposited Campanian Ignimbrite tephra layer allows a straightforward correlation to the Urluia record. The similar patterns in the χfd and especially

217

grain-size from Vlasca and Urluia allow a convincing correlation of these two sections, and thus, an indirect correlation of the Vlasca sequence to the Greenland δ18O record. Furthermore, the Vlasca and Urluia records were compared with the close-by Rasova section providing an age-depth model (Zeeden et al., in press), supporting and additionally verifying the presented age models (Fig.S7.6).

The onset of the studied part of the Urluia section is tentatively correlated to Greenland

Stadial (GS) 15 (before 55.4 ka; Fig. S7.4). Following an increase in χfd, high values at 8.06 m are corresponding to Greenland Interstadial (GI) 14. The following points of correlation are associated with increases in χfd at 7.58 m, 7.18 m, and 6.70 m, and are related to GI 13, GI 12, and GI 11 (Fig. S7.4), respectively. The luminescence sample C-L3708 at 6.90 m indicates a timing of deposition of this layer is 54.2+/-4.1 with 1-σ uncertainty, or 54.2+/-8.0 with 2-σ uncertainty (Fig. S7.2). The correlative age (~45 ka for 6.90 m) is in the lower range of the 2- σ uncertainty. However, as noted above, it is possible that luminescence ages tend to show an age in a lower range of the 2-σ uncertainty.

Establishing the onset of the Vlasca profile is very challenging since the grain-size is heavily influenced by high sand content that is very likely related to a local flooding event and therefore regarded as a non-climatic signal. The base of the section is deposited as a sandy layer. Generally high χfd and sand content in these layers (Fig. 7.2) indicate that such grain- size distribution is not a climatic signal, but rather related to the increased fluvial dynamics or flooding of the Danube River that supplied the Danube banks with high amounts of sand and coarse particles suitable for aaeolian transport. We related the onset of the Vlasca section to

GS 12, while the increases in χfd at 10.08 m and 9.60 m are related to GI 11 and GI 10 (Fig. S7.5). However, such correlation for the older part of Vlasca section is highly debatable, and should be treated with special caution.

The preserved tephra layer is a safe chronological marker and it is used as a reliable tie point at both sections with an age of 39.93 ± 0.1 ka (Fitzsimmons et al., 2013; Lowe et al., 2012;

Veres et al., 2013). Above the tephra layer, χfd and fine grain-size fractions at both sections indicate a prominent peak in data, which we relate to Greenland Interstadial (GI) 8 (Figs. S7.4 and S7.5). At Urluia, the luminescence sample is present in this depth interval (Fig. S7.2). As mentioned above, it exhibits an age of 44.1 ± 3.4 ka (with 1-σ uncertainty), a clearly older age than the deposition of Campanian Ignimbrite tephra. However, the timing of the GI 8 at

218

38,300 years ago is in the range of the luminescence age uncertainty. The enhanced peak in

χfd and finer grain-size fraction are also observed at Vlasca, and this layer is related to GI 8 as well.

At Vlasca, three prominent peaks in χfd and finer grain-size fraction stands out at ~8.2 m, ~7.86 m and ~7.52 m (Fig. S7.5). These layers were correlated to GI 7, GI 6 and GI 5. At Urluia, only two clear peaks in the fine fractions can be observed. Two luminescence ages at 5.20 m and 4.50 m support the correlation of sediment interval from 4-6 m to tentatively span the timeframe of the past 30,000 to 40,000 years (Fig. S7.2). The luminescence age at 4.5 m (32.5 ± 2.5 ka; C-L3712) suggests a correlation of the younger peak in grain-size data to the GI 5. However, the luminescence age of 36.2 ±2.8 ka (C-L3711) cannot clearly specify the increase in fine fractions to GI 6 or GI 7. We argue that this increase in finer fractions corresponds to GI 6, while GI 7 is recorded at a depth of 5.62 m. Two main criteria for such correlation are 1) the general tendency of the here presented luminescence ages to indicate slightly older ages or expected ages to be in the younger uncertainty limit and 2) observed similarities to the Vlasca section. Regarding the first criterion, the luminescence age that dates the sediment just below the observed increase in fine particles would support a correlation with 1-σ uncertainty of this layer to both, GI 6 and GI 7. However, considering that the means of the older luminescence ages are slightly overestimated, especially indicated by the Campanian Ignimbrite, it is more likely that the lower limits of the ages uncertainties gives a better assumption of the ‘true’ age. Therefore, we correlated the peak of higher contribution of fine particles at 5.02 m to GI 6, while the relatively weaker increase at 5.62 m in χfd and grain-size we relate to GI 7 (Fig. S7.4). Generally higher χfd and slightly finer particles at 5.62 m than at 5.02 m are in better agreement with the values related to GI 7 at Vlasca.

Establishing the following two tie points of the presented age model was generally straightforward, since the χfd indicate a notable increase in the values, while the U-ratio indicate a decrease. At Urluia, the layers at 3.60 m and 3.32 m are related to GI 4 and GI 3, while at Vlasca these interstadials are related to depths of 6.50 m and 6.22 m (Figs. S7.4 and

S7.5). However, although the χfd indicates a clear increase in humidity, the grain-size distribution shows a higher contribution of coarser fractions (Figs. S7.4 and S7.5). Therefore, it may be debated if this layer showing a high contribution of coarse particles at both sections corresponds to Heinrich event (HE) 3. Nonetheless, we argue that only part of the observed increase in grain-size is related to HE 3, where the coarser particles are a consequence of 219

increased winter severity as recorded in the Black Sea record (Wegwerth et al., 2015a) that covers GS 5 and 4 (Fig. S7.7). In addition, the luminescence age at 3.3 m (C-L3713) supports the established correlation (Fig. S7.2).

An increase in χfd and finer grain-size at 5.16 m at the Vlasca section is correlated to GI 2

(Fig. S7.5). Although similar increase in χfd cannot be observed at Urluia, but a clearly similar pattern in grain-size decrease is observed at 2.68 m and correlated to GI 2 (Fig. S7.4). Nevertheless, such correlation is not strongly supported by luminescence age at 2 m (25.0±1.9 ka, C-L3715). The youngest uncertainty of this age is ~23 ka, indicating the timing of the GI 2 somewhere above 2 m at Urluia. However, a correlation based on this luminescence age would indicate very low sedimentation rates for last 23 ka, which includes most of the Last Glacial Maximum. Although we cannot rule this scenario out, we assume it as unlikely.

Therefore, we correlate the increase in grain-size fractions and slight decrease in χfd at 2.30 m to GS 3. Based on the relation of χfd and grain-size from Urluia and Vlasca, we related the interval at 4.02 m at Vlasca to GS 3 (Figs. S7.4 and S7.5).

The luminescence age suggests for the sediment at 1.5 m depth a lower uncertainty age of ~19.5 ka (Fig. S7.2). According to that age, the sediment above showing a decrease in the <5

µm fractions and χfd at 1.30 m is related to HE 1. At Vlasca, HE 1 is correlated to the depth of 2.54 m. Such correlation is in good agreement with a correlation obtained at the Rasova section (Zeeden et al., in press), where the HE 1 is recorded shortly after three prominent peaks in χ that can be also observed at the here discussed sections (Fig. S7.6).

For the timing of the Younger Dryas we related a decrease in χfd and an increase in grain-size fractions after the second prominent peak in those proxies observed in the younger part of sections to the Younger Dryas. We argue that those lower values are generally unlikely to represent the Holocene climate. A following increase in χfd and fine grain-size is related to the onset of the Holocene. The same solution was obtained for the Rasova section (Zeeden et al., in press), which makes a regional correlation more coherent (Fig. S7.6).

220

Table S7.4. Tuning points of the Urluia and Vlasca sections correlation to δ18O from Greenland ice sheet (Andersen et al., 2004)

Profile Tuning point Depth (m) Age (years) Urluia GS 15 8.48 55,400 Urluia GI 14 8.06 54,800 Urluia GI 13 7.58 49,800 Urluia GI 12 7.18 47,350 Urluia GI 11 6.70 43,400 Urluia Campanian 6.22 39,930 Ignimbrite tephra Urluia GI 8 5.90 38,300 Urluia GI 7 5.62 35,300 Urluia GI 6 5.02 33,500 Urluia GI 5 4.38 32,250 Urluia GI 4 3.60 28,500 Urluia GI 3 3.32 27.400 Urluia GI 2 2.68 22,700 Urluia GS 3 2.30 20,450 Urluia HE 1 1.46 15,450 Urluia Holocene transition 0.88 11,500 Vlasca GS 12 10.72 44,000 Vlasca GI 11 10.06 43,400 Vlasca GI 10 9.60 41,750 Vlasca Campanian 9.12 39,930 Ignimbrite tephra Vlasca GI 8 8.58 38,300 Vlasca GI 7 8.20 35,300 Vlasca GI 6 7.86 33,500 Vlasca GI 5 7.52 32,250 Vlasca GI 4 6.50 28,500 Vlasca GI 3 6.20 27.400 Vlasca GI 2 5.16 22,700 Vlasca GS 3 4.02 20,450 Vlasca HE 1 2.54 15,450 Vlasca Holocene transition 1.52 11,500

221

Figure S7.4. Correlation between the δ18O data from Greenland ice sheet (Andersen et al.,

2004) (blue line) and proxies from the Urluia section (green lines for χfd (see the discussion in chapter 4.2 for the explanation of two different scales), purple line for U-ratio, red line for fine particles (<5 µm) and orange line represents mean grain-size ).

222

Figure S7.5. Correlation between the δ18O data from Greenland ice sheet (Andersen et al.,

2004) (blue line) and proxies from the Vlasca section (green line for χfd, purple line for U- ratio, red line for fine particles (<5 µm) and orange line for mean grain-size). Grey rectangle represents a sandy layer that is not considered for palaeoclimate reconstruction (not shown elsewhere).

223

Figure S7.6. Correlation between the δ18O data from Greenland ice sheet (Andersen et al.,

2004) (blue line), and χ and χfd from Vlasca (orange and red lines), Urluia (lines in different shades of green; see chapter 4.2 for the explanation of two different scales) and Rasova (Zeeden et al., in press) (lines in different shades of purple) sections presented on the correlative age models. Grey rectangles represent Heinrich Events, straight yellow line represent the timing of the Campanian Ignimbrite tephra deposition.

224

Figure S7.7. Comparison of mean grain-size from Vlasca (blue line) and Urluia (green line), and the accumulation rate of coastal ice-rafted detritus (IRDC) from the Black Sea as an indication of winter severity (Wegwerth et al., 2015a, 2015b).

225

4.2.3. Final age model

To obtain the final age model we have combined a correlative model and luminescence ages and used the most certain tie points only. The Campanian Ignimbrite tephra is used as a robust tie point at both sections. The following two tie points at Urluia are at 4.5 m (32.5 ka) and 3.3 m (28.7 ka), where luminescence samples show very good relation to the correlative age model (Fig. S7.2). The next tie point is at 0.9 m (11.5 ka), where an increase in all proxies indicates the onset of the Holocene. Since the Vlasca section has no further time control, we relate obvious changes in χfd which can be observed at both sections to obtain the same age.

The low value in χfd at 7.38 m at Vlasca is related to the low value at 4.08 m at Urluia and assumed to have an age of 31,170 years (Figs. 7.2, S7.4 and S7.5). An increase in χfd at 6.22 m is related to an increase at 3.32 m at Urluia, and has an assumed age of ~28,760 years (Figs.

7.2, S7.4 and S7.5). An increase in χfd at 2.1 m and 1.52 m at Vlasca are related to increases at 1.32 m (~14,630 years) and 0.88 m (11.500 years) at Urluia (Figs. 7.2, S7.4 and S7.5). Table S7.5 shows tie points for both sections.

Finally, it is important to state that no matter which age model is used, the conclusions presented in this study are valid for all three age models (Fig. S7.8). The difference to data from the Middle Danube Basin is also evident when no age model additional to the Campanian Ignimbrite occurrence is applied.

Table S7.5. Tie points of the final age models for the Urluia and Vlasca sections.

Profile Depth [m] Age [years] Urluia 8.48 55,400 Urluia 6.22 39,930 Urluia 4.50 32,500 Urluia 3.30 28,700 Urluia 0.88 11,500 Vlasca 10.00 43,000 Vlasca 9.12 39,930 Vlasca 7.38 31,170 Vlasca 6.20 28,760 Vlasca 2.10 14,630 Vlasca 1.52 11,500

226

Figure S7.8. Comparison of the correlative age model (red lines) and final age model (black lines). Note the same general trends regardless which age model is used.

227

Figure S7.9. Major elements from the Urluia and Vlasca sections. Major elements are presented in %.

228

Figure S7.10. Geochemical ratios commonly used to indicate possible changes in the provenance. No ratio indicates major change in provenance, only Al2O3/K2O ratio at Vlasca suggests a slight increase in K-feldspars between 20-10 ka.

229

Figure S7.11. Frequency dependent magnetic susceptibility from Rasova (Zeeden et al., in press) (red line), Vlasca (blue line) and Urluia (green line), U-ratio from Vlasca (pale blue line) and Urluia (dark green line) compared to Land Evolution Zones (LEZ) form laminated Eifel maar sediments (Sirocko et al., 2016) of the last 60,000 years. Note the similar timing of the transition of LEZ 7 (boreal forest) to 6 (steppe) and continentalization of the Lower Danube Basin.

230