Single-Grain Zircon U-Pb Dating and Magnetic Susceptibility of Polish Loess to Determine Late Quaternary Dust Provenance and Paleoclimate

Examensarbete vid Institutionen för geovetenskaper

Degree Project at the Department of Earth Sciences

ISSN 1650-6553 Nr 446

Single-Grain Zircon U-Pb Dating and Magnetic Susceptibility of Polish Loess to Determine Late Quaternary Dust Provenance and Paleoclimate

U-Pb datering av zirkoner och magnetisk susceptibilitet i polsk lössjord för fastställning av kvartära härkomstområden och paleo- klimatologi av atmosfäriskt damm

Alexandra Engström Johansson

INSTITUTIONEN FÖR GEOVETENSKAPER

D E P A R T M E N T O F E A R T H S C I E N C E S

Examensarbete vid Institutionen för geovetenskaper

Degree Project at the Department of Earth Sciences

ISSN 1650-6553 Nr 446

Single-Grain Zircon U-Pb Dating and Magnetic Susceptibility of Polish Loess to Determine Late Quaternary Dust Provenance and Paleoclimate

U-Pb datering av zirkoner och magnetisk susceptibilitet i polsk lössjord för fastställning av kvartära härkomstområden och paleo- klimatologi av atmosfäriskt damm

Alexandra Engström Johansson

ISSN 1650-6553

Copyright © Alexandra Engström Johansson

Published at Department of Earth Sciences, Uppsala University (www.geo.uu.se), Uppsala, 2018

Abstract

Single-Grain Zircon U-Pb Dating and Magnetic Susceptibility of Polish Loess to Determine Late Quaternary Dust Provenance and Paleoclimate

Alexandra Engström Johansson

Windblown mineral dust has accumulated into thick loess layers across the European continent. Loess deposits are valuable paleoclimatic archives as they record climate change on centennial- to glacial- interglacial timescales. A reliable identification of dust sources is of fundamental importance for paleoclimatic reconstructions of atmospheric circulation patterns and dust pathways, and for understanding the link between mechanisms of dust production and -emissions and climate change.

European dust sources are however still relatively poorly constrained, which in part is related to the few provenance studies performed using single-grain techniques. Single-grain techniques, and in particular single-grain U-Pb dating of detrital zircons, may allow for a reliable identification of multiple individual sources, as has been shown for the Chinese loess archive. Few dust provenance studies in general, and none found using single-grain techniques, have been performed on Polish loess samples.

The location of Polish loess deposits in the central parts of the European continent may allow for a more complete paleoclimatic picture by tying together provenance studies from west to east, as well as for distinguishing between the proposed central European dust source areas of distal Fennoscandia and the proximal mountains of Moravia, the Alps, the Sudetes, the Bohemian Massif and the Carpathians.

Forty-two loess samples were collected from the Biały Kościół loess profile in southwestern Poland, spanning deposition ages from the Eemian to the Holocene (MIS 5-1). Five samples were dated using single-grain zircon U-Pb geochronology and thirty-seven samples were measured and analysed using magnetic parameters of mass-specific-, frequency- and phase dependent susceptibility which reflect changes in mineralogy and grain size of magnetic mineral fractions.

Magnetic susceptibility results show variations between and within stratigraphic units. Different soil formation processes appear to have been dominant during the three paleosol units (Holocene, MIS 3, MIS 5), while variability in magnetic susceptibility during MIS 4 and 2 may be related to source changes caused by changes in climatic conditions. Single-grain zircon U-Pb age distribution data indicate similar dust provenance for the Biały Kościół loess since the Eemian, with a very prominent age peak at ca. 325 Ma that is reflected to varying degree in all other sediment samples. The prominence and relative importance of the 325 Ma peak is best matched by sites along the lower Danube and the Tisza River, and corresponds to crustal rock ages related to the Variscan orogeny. Variscan source terranes can be found across central Europe and may suggest the outer Western Carpathians and the Bohemian Massif as potential primary source rocks for the Biały Kościół loess.

Keywords: dust provenance, Polish loess, paleoclimate, single-grain zircon U-Pb dating, magnetic susceptibility

Degree Project E1 in Earth Science, 1GV025, 30 credits
 Supervisor: Thomas Stevens 


Department of Earth Sciences, Uppsala University, Villavägen 16, SE-752 36 Uppsala (www.geo.uu.se) ISSN 1650-6553, Examensarbete vid Institutionen för geovetenskaper, No. 446, 2018

The whole document is available at www.diva-portal.org

Populärvetenskaplig sammanfattning

U-Pb datering av zirkoner och magnetisk susceptibilitet i polsk lössjord för fastställning av kvartära härkomstområden och paleoklimatologi av atmosfäriskt damm

Alexandra Engström Johansson

Atmosfären är full av små partiklar som bildats genom nedbrytning av berg och sten. Vinden kan transportera detta atmosfäriska stoft tusentals kilometer från sina härkomstområden innan det faller till marken och landar på våra kontinenter. Med tiden har det bildats tjocka jordlager, så kallade lössjordar, i många av Europas länder. Vetenskapliga undersökningar visar att förändringar i klimat, såsom mellan istider, kan kopplas till förändringar i lössjordarnas kemiska sammansättning. Lössjordarna kan därmed ses som ett arkiv över Europas klimatförändringar. Om vi kan tolka den värdefulla information som dessa klimatarkiv gömmer så kan vi försöka rekonstruera Europas klimathistoria och bättre förstå kopplingen mellan atmosfäriskt damm och klimatförändringar. Då kan vi också mer pålitligt förutspå hur framtida klimat kan komma att se ut.

Det fattas dock fortfarande viktig information om ursprunget av Europas stoft. U-Pb datering av korn av det hårda mineralet zirkon är en metod som har använts för att identifiera de bergarter som är ursprunget till kinesiskt damm, och som bidragit till att vi har en bättre förståelse för hur lössjordar i Kina har bildats. Få studier med denna metod har däremot gjorts för europeiska lössjordar, och inga för polsk löss. Tidigare studier har föreslagit att stoft i centrala Europa härstammar antingen från den avlägsna fennoskandiska bergskedjan eller från de mer närliggande bergsområdena i Mähren, Alperna, Sudeterna, Böhmen och Karpaterna. Polens centrala placering i Europa gör landet till en idealisk plats att undersöka och jämföra den påverkan som avlägsna vs närliggande potentiella stoft-källor haft på europeiska lössjordar.

Fyrtiotvå lössprover togs från Biały Kościół i sydvästra Polen, motsvarande åldrar från ca 100 000 år sedan till moderna klimatförhållanden. Fem prover analyserades med U-Pb datering av zirkoner och trettiosju prover analyserades genom att mäta olika egenskaper av så kallad magnetisk susceptibilitet.

Magnetisk susceptibilitet påverkas av förändringar i sammansättningen och kornstorleken av magnetiska mineraler.

Resultaten visar att magnetisk susceptibilitet varierar mellan prover av olika åldrar, vilket kan vara orsakat av förändringar över tid i klimatförhållanden och av de områden som producerar det atmosfäriska stoftet. Resultatet från U-Pb dateringen av zirkoner visar att ålderssammansättningen av lössjorden från Biały Kościół är densamma för alla fem prover, vilket tyder på att källorna till stoftet inte förändrats över tid. Den största andelen zirkon-korn har en ålder av ca 325 miljoner år. Bergarter av denna ålder kan kopplas till bildandet av den Varisiska bergskedjan och återfinns i områden i centrala Europa och i de närliggande västra Karpaterna. Dessa kan därmed vara en möjlig källa till det polska stoftet.

Nyckelord: härkomstområden för atmosfäriskt stoft, polsk lössjord, paleoklimat, U-Pb datering av zirkoner, magnetisk susceptibilitet

Examensarbete E1 i geovetenskap, 1GV025, 30 hp
 Handledare: Thomas Stevens

Institutionen för geovetenskaper, Uppsala universitet, Villavägen 16, 752 36 Uppsala (www.geo.uu.se) ISSN 1650-6553, Examensarbete vid Institutionen för geovetenskaper, Nr 446, 2018

Hela publikationen finns tillgänglig på www.diva-portal.org

Table of Contents

1

Introduction ... 1

2

Central European loess ... 6

2.1 Chemical composition, grain size and influencing factors ... 6

2.2 Low field magnetic susceptibility ... 7

2.2.1 Frequency dependent susceptibility ... 11

2.2.2 Out-of-phase susceptibility ... 13

2.3 Dust provenance analysis ... 15

2.3.1 Single-grain zircon U-Pb dating ... 15

2.3.2 Dust sources, -emissions and paleoclimate ... 18

2.3.3 Primary and secondary dust sources ... 21

3 Materials and methods ... 24

3.1 Study site and sampling ... 24

3.2 Magnetic susceptibility ... 26

4 Results ... 30

4.1 Magnetic susceptibility ... 30

4.2 U-Pb ages of detrital zircons ... 34

5 Discussion ... 39

5.1 Magnetic susceptibility and paleoclimate ... 39

5.2 Polish loess dust sources ... 41

6 Conclusions ... 45

7 Acknowledgements ... 46

8 References ... 47

9 Appendix ... 55

Appendix I: Zircon U-Pb Supplementary Data PO1 ... 55

Appendix II: Zircon U-Pb Supplementary Data PO2 ... 59

Appendix III: Zircon U-Pb Supplementary Data PO3 ... 63

Appendix IV: Zircon U-Pb Supplementary Data PO4 ... 67

Appendix V: Zircon U-Pb Supplementary Data PO5 ... 71

1

1 Introduction

Atmospheric mineral dust (hereafter called dust) emissions, amounts and pathways are affected by changes in climate parameters such as aridity, temperature and atmospheric circulation patterns. In turn, windblown dust influences global climate by affecting the Earth’s radiative budget, patterns of atmospheric circulation and precipitation, and biogeochemical cycles (Maher et al., 2010; Choobari et al., 2004; Harrison et al., 2001). Quaternary dust variations recorded in ice cores and sedimentary archives have been shown to correlate with variations in paleoclimatic indicators (Ruth et al., 2003), such as the oxygen isotope record and atmospheric CO

concentrations, indicating a link between dust flux and climate variability (Újvári et al., 2017; Moine et al., 2017; Harrison et al., 2001; Kukla, 1970). An understanding of the dynamic and complex feedbacks that govern the interactions between dust and climate change is therefore of fundamental importance for reconstructions of past climate conditions, which in turn are the basis for reliable projections of future climate (Knippertz & Stuut, 2014; Arimoto, 2001). Decades of dust research and various recent technological advances have dramatically improved our understanding of dust dynamics and effects (Muhs, 2013), but uncertainties still remain regarding causes of dust emissions, amounts of dust in the atmosphere (Maher et al., 2010), and the connection to abrupt climate change on centennial- to millennial timescales (Újvári et al., 2017; Rousseau & Sima, 2014, Rousseau et al., 2013). This is in part related to the lack of knowledge of dust sources (Rousseau et al., 2014).

Dust from different source areas mix in the atmosphere and is subsequently deposited up to thousands of kilometres from its place of origin, over time accumulating into thick sediment layers of loess that extend across vast areas of central and eastern Europe (Fig.1)(Haase et al., 2007; Pye, 1995). The northernmost loess “belt” stretches from Great Britain and France in the west, across Germany, the Czech Republic, Slovakia and Poland to Ukraine and Russia in the east. These loess deposits include interbedded layers of paleosols, “fossil” soils preserved through burial. Loess-paleosol sequence stratigraphy has been shown to correlate with Quaternary glacial-interglacial cycles, as well as to record climate shifts on centennial to millennial timescales (Újvári et al., 2017; Lindner et al., 2002; Rousseau et al., 2002; An &

Porter, 1997; Kukla, 1970). The loess layers are thought to reflect the cold and arid conditions,

strong winds and high accumulation rates of glacial and stadial periods. The paleosols are

thought to reflect the warmer and/or more humid climate conditions and reduced accumulation

rates characteristic of interglacial and interstadial periods, when increased weathering rates

2

allowed for the development of soils (Kemp, 2001). Magnetic susceptibility measures the amount of pedogenic iron oxides formed in loess as an effect of weathering (Heller & Evans, 1995). Using magnetic susceptibility as a climate proxy, glacial-interglacial cycles as well as climate variability on shorter timescales have been identified in both the Chinese and the European terrestrial loess archives and correlated to the oxygen isotope record in deep-sea sedimentary records (Kukla, 1970; 1988). European loess deposits may therefore contain key information for understanding both continental and global climate evolution (Markóvic et al., 2015; Muhs, 2013, Buggle et al., 2009).

Several different processes have been proposed to have driven the production of dust in Europe during the Quaternary, including changes in: sea level, ice sheet and glacier extent, river system development, atmospheric circulation patterns and aridity (Obreht et al, 2016; Rousseau et al., 2014; Újvári et al., 2012; Smalley, 1995). Each proposed process is caused by a distinct set of production mechanisms. Identifying the right one is crucial for a correct interpretation of how climate controls and responds to dust flux variability. Knowledge of dust provenance may allow for a differentiation between possible production mechanisms, as geographical factors (e.g. proximity to the ocean and to ice sheet- and glacier margins) determine the relative potential influence of each process. Reliably identifying dust source areas could therefore contribute to an increased understanding of the connection between dust and climate change (Stevens et al., 2010; Harrison et al., 2001).

Knowledge of European dust provenance is however limited. Earlier studies used mineralogical and geochemical (major and trace elements) analyses of bulk samples to identify possible source areas. Loess from Ukraine is suggested to have been derived from local Fennoscandian glaciofluvial deposits, while Hungarian and Serbian loess is proposed to derive from alluvial fans of the Danube river formed by input from local felsic sources (possibly the Austrian Alps and the Carpathians). Romanian loess is suggested to be a mixture of both (Buggle et al, 2008, Ùjvári et al., 2008; 2014).

While results from bulk samples may provide a general understanding of dust provenance,

it does not easily allow for differentiating between multiple individual sources. Single-grain U-

Pb dating of detrital zircons is a technique better suited for this purpose as it investigates

individual grains one at a time, and so can trace multiple individual sources. It has been

successfully used in provenance analyses of Asian dust, and as such has been of great

importance for understanding the evolution of Chinese loess archives (Fenn et al., 2018; Licht

et al., 2016; Nie et al., 2015; Stevens et al., 2013).

3

Initial results using single-grain detrital zircon U-Pb dating on late Quaternary European loess deposits point to: i) the Bohemian Massif and the Eastern Alps as dust sources for Austrian loess (Újvári et al., 2013; 2015); ii) the Alps, the Carpathians and local highlands as dust sources for Hungarian Danube loess (Újvári et al., 2012), and iii) the Carpathians and adjacent local highlands as dust source areas for Ukrainian loess (Nawrocki et al., 2018). Initial fluvial transport is believed to have been responsible for a significant part of grain entrainment, transport and accumulation before subsequent aeolian reworking (Nawrocki et al., 2018; Újvári et al., 2012; 2013; 2015). Attempts have also been made to identify European loess sources using atmospheric dust cycle modelling (Rousseau et al., 2014). A reliable distinction between proposed dust sources still remains difficult however, as does constraining their relative influence and spatial and temporal resolution. These uncertainties, and the overall lack of single-grain provenance studies, affect the validity of climate proxy interpretations in loess as dust provenance is a strong influencing factor (Knippertz & Stuut, 2014).

Poland’s geographical location in the central parts of Europe places the country in an ideal place to investigate the relative influence of the proposed proximal sources of the Carpathian Mountains, Eastern Alps, the Sudetes, Moravia and the Bohemian Massif on central European dust sources, as well as to investigate the role of northern glacial debris from the more distal Fennoscandian Shield. Studies of Polish loess archives could also allow for a “tying together”

of loess provenance studies from west to east, for a more complete understanding of the continents paleoclimatic archives (Jary & Ciszek, 2013). A few studies have attempted to address the question of Polish dust sources. Smalley and Leach (1978) thought it obvious that Polish loess was to be primarily considered a product of glacial grinding by the northern glaciers before subsequent deposition by aeolian deflation. This hypothesis was revised by Badura et al.

(2013) who proposed major pleniglacial paleoriver transport as the initial supplier of silty material from source areas in the Sudeten Mountains and Moravia (Fig.1) before subsequent aeolian reworking. Micromorphological evidence of older, pedogenic modification in younger Polish loess published by Mroczek et al. (2013) suggests episodes of soil formation at a previous location, which supports the idea of aeolian processes being a second mode of transport for late Quaternary deposits.

These studies are however based on interpretations of sediment stratigraphy, grain

morphology and landscape geomorphology, and do not provide sufficient evidence for a

reliable identification of dust sources. There is therefore high potential for gaining new

information of central European dust provenance by using single-grain zircon U-Pb dating to

compare and match zircon age populations in Polish loess with that of possible source areas.

4

Adding another piece to the puzzle of European dust sources may provide valuable information for the interpretation of dust production mechanisms, driving forces for dust emissions and the connection between cause and effect in the interaction between dust and climate change.

In this thesis I intend to investigate and identify late Quaternary dust sources of loess from the site Biały Kościół (Fig.1) in southwest Poland using single-grain zircon U-Pb dating.

Magnetic susceptibility is used to construct a rudimentary weathering profile for confirmation

of stratigraphic age boundaries and to place the U-Pb ages into a temporal and paleoclimatic

context. In addition to low-field magnetic susceptibility, I will investigate the magnetic

parameters of frequency- and phase dependence that may yield granulometric information

useful for paleoclimatic interpretations (Hrouda et al., 2011; 2013). The overall aim of this

study is to use the results of loess magnetic susceptibility and dust provenance as the basis for

a discussion on climate controls on dust emissions and implications for central European

paleoclimate since the end of the Eemian (the previous interglacial).

5

Figure 1. Map of the spatial distribution of European loess deposits. Red star marks the study site of Biały Kościół in SW Poland. Smaller abbreviations in black are city names of major Eurasian cities (Haase et al., 2007).

6

2 Central European loess

2.1 Chemical composition, grain size and influencing factors

Approximately 1/5 of the surface area of Europe is covered by loess, making loess the parent material for much of the continent’s soils. Two main areas of loess are distributed across central Europe: i) a continuous belt (50-100 km wide) north of Europe’s mountainous regions and; ii) the northern foreland of the Alps along the central and lower parts of the Danube river (Fig.1)(Haase et al., 2007).

Loess mainly consists of silt- and clay-sized rock and mineral particles, composed of the major elements of the main rock-forming minerals; Si, Al, O, Fe, Ca, K, Na, Mg, Mn, Ti and P (Sheldon & Tabor, 2009). Silicates (e.g. quartz, feldspars, clay minerals) generally dominate the mineralogy together with varying amounts of carbonates, sulfates and Fe-bearing minerals (Knippertz & Stuut, 2014). Accessory minerals include the heavy mineral zircon (ZrSiO

) which, due to its highly resistant properties, is concentrated by sorting processes and ubiquitously found in loess (Hawkesworth & Kemp, 2006). Zircons are considered the best provider of the isotope measurements used in U-Pb geochronology and are therefore of great value for dust provenance studies, as they are highly resistant to chemical- and physical weathering (and therefore thought to date crystallizations events) and found ubiquitously in felsic igneous rocks and siliclastic sediments (Hawkesworth & Kemp, 2006; Fedo et al., 2003).

The physiochemical properties of loess reflect the physiochemical properties of sediments in dust source areas and could therefore be used to as source indicators. The complex modification processes that act on the geochemical composition and grain size of dust during entrainment, transport and after deposition (e.g. grain size sorting, admixing of non-aeolian sediments, mechanical disaggregation and abrasion, climate, bioturbation), do however complicate the use of these parameters as provenance proxies as loess deposits may no longer adequately reflect dust source rock properties (Knippertz & Stuut, 2014; Kemp, 2001;

Johnsson, 1993).

Increased weathering rates (due to higher temperatures, higher precipitation rates, lower

accumulation rates) alters loess geochemistry by i) removing soluble elements (e.g. Na, Ca,

Mg), ii) concentrating immobile elements (e.g. Al, Ti, Si), and iii) producing secondary clay

minerals (e.g. illite, kaolinite), secondary carbonates, nodules of iron and manganese, and

ultrafine pedogenic iron oxides (Buggle et al., 2011; Kemp, 2001; Heller & Evans, 1995).

7

Changes in wind strength affects the average grain size of loess, as stronger winds have the capacity to entrain coarser, heavier grains compared to weaker winds. A change in dust source also affects average loess grain size as this changes the distance of travel between source area and place of deposition. A shift towards a more distal (proximal) source area would reduce (increase) the average grain size for an unchanged wind direction and strength, as lighter smaller grains can travel further than heavier coarser grains (see Fig. 2 for dust transportation mechanisms)(Licht et al., 2016; Liang et al., 2013; Prins et al., 2007). Changes in grain size also affect the chemical composition of loess, as different elements are enriched in different grain size fractions (Liang et al., 2013; Buggle et al., 2011).

Thus, the cause of variations in the physiochemical composition of loess is not easily attributed to a change in any one parameter and may well be the result of the interplay between changes in more than one parameter (Kemp, 2001). Using the physiochemical properties of loess to identify changes in dust provenance is therefore a complicated task, one that requires a good understanding of the processes involved (Buggle et al., 2011; Kemp et al., 2001).

Figure 2. Dust transport- and deposition mechanisms from distal and proximal sources areas (Revised from Pye, 1995).

2.2 Low field magnetic susceptibility

Low field (or initial) magnetic susceptibility (hereafter referred to as magnetic susceptibility or

simply susceptibility, denoted 

) measures the degree to which a rock, sediment or

environmental material can be temporarily “magnetized” when subjected to a known, external

weak (low) magnetic AC-field (alternating field)(Heller & Evans, 1995; An et al., 1991; Kukla

et al., 1988).

8

K

is a dimensionless parameter, defined as the acquired magnetization (M) per unit magnetic field (H)(Evans & Heller, 2003):

 = M/H

All materials have a fundamental property of diamagnetism (the potential for temporary magnetization by an external magnetic field), including many naturally occurring minerals (e.g.

quartz and calcite) and water, although the effects of diamagnetism on the susceptibility signal are generally negligibly small (Evans & Heller, 2003; Moskowitz, 1991). The strength of the magnetic susceptibility signal is generally determined by the composition, concentration and grain size of Fe-bearing minerals (Buggle et al., 2009; Evans & Heller, 2003; Kukla et al., 1988). These can be categorized by their different magnetic properties into para-, ferro-, antiferro- and ferrimagnetic minerals. The magnetization, and thus the susceptibility, of ferro- (e.g. Fe and many of its alloys) and ferrimagnetic minerals (e.g. magnetite, maghemite) is much stronger than both antiferro- (e.g. hematite) and paramagnetic (e.g. biotite, pyrite) minerals (Buggle et al., 2009; Moskowitz, 1991). Changes in concentrations of the former would therefore have a disproportional effect on the susceptibility signal compared to the latter (Buggle et al., 2009).

Although the exact production mechanism is still not understood, it is generally accepted the that magnetic susceptibility in loess measures the concentration of ultrafine iron oxides (ferrimagnets) that can form in situ under soil-forming conditions (Maher & Taylor, 1998;

Maher & Thompson, 1991). The main carriers are fine-grained (<1m) magnetite (Fe

O

) and maghemite (Fe

O

)(Kukla et al., 1988; Maher & Thompson, 1991). Hematite and goethite may also influence magnetic susceptibility values, but to a significantly lower degree (Buggle et al., 2009). These ultrafine pedogenic iron-oxides (~0.018-0.020 m) are sufficiently small to have superparamagnetic (SP) properties that allows for an instant response to any applied magnetic field. This magnetic mineral grain size fraction therefore has the highest susceptibility (Buggle et al., 2009; Evans & Heller, 2003). As grain size increases, particles acquire single-domain (SD) and multi-domain (MD) properties. The exact grain size threshold between the different categories depends on the type of mineral in question, but generally there is an increased susceptibility with reduced grain size (Coey, 2010; Buggle et al., 2009; Evans & Heller, 2003).

Magnetic susceptibility has long been an established indicator of climate change in loess

deposits and has been widely used in stratigraphic studies to distinguish between loess and

9

paleosols, and for correlation with other climate archives and geochronological frameworks, such as the oxygen isotope record of marine sediments (Fig. 3)(Hosek et a., 2015; Buggle et al., 2008; 2009; Porter et al., 2001; Nawrocki et al., 1996; 1999; An et al., 1991; Kukla et al., 1988). Readily available portable devices have allowed for high-resolution measurements to be performed quick and easy in the field (Maher, 1998; Heller & Evans, 1995).

High susceptibility values in paleosols (dominated by SP-grains) and low in loess (dominated by SD- and MD-particles) have been measured across Europe, reflecting enhanced weathering conditions, lower accumulation rates and thus increased pedogenic activity during interglacials compared to glacials (Hosek et al., 2015; Antoine et al., 2013; Bokhorst et al., 2009; Buggle et al., 2008; 2009; Evans & Heller, 2003). Low field magnetic susceptibility has therefore been found a useful weathering proxy in central European loess deposits from the Czech Republic, Poland and Ukraine (Hosek et al., 2015; Antoine et al., 2013; Bokhorst et al., 2009; Buggle et al., 2008; 2009; Nawrocki et al., 1996; 1999).

The technique has been used to identify “non-weathered” loess in which the susceptibility signal is thought to be an indicator of “background susceptibility”, reflecting the composition and concentration of detrital magnetic minerals and weathering in source regions rather than post-depositional alteration (Buggle et al., 2008; 2009).

Figure 3. A comparison between lithology (L=loess, S=soil) and low field magnetic susceptibility (S) of the two Chinese loess sites of Xifeng and Luochuan and the deep-sea oxygen-isotope chronology of Prell et al. (1986) and Imbrie et al. (1984). Black circles correlate tops of oxygen isotope stages with their suggested loess age equivalent.

Arrows indicate the Brunhes/Matuyama paleomagnetic boundary at

~

10

Variations in “pure” loess susceptibility values within a site could then be caused by local effects on: i) degree of weathering in source regions; ii) degree of weathering during accumulation; iii) magnetic mineral content and grain size related to source changes (Buggle et al., 2008). Local variations in susceptibility within paleosols can occur when post-depositional processes of leaching and precipitation differentially affect soil horizons, removing magnetic minerals in one (lowering susceptibility) and increasing them in another (increasing susceptibility)(Hosek et al., 2015).

Magnetic susceptibility has been found to be less reliable as a weathering proxy in loess deposits linked to wetter climatic conditions, such as periglacial environments (Fig. 4), in some central European localities (Hosek et al., 2015; Antoine et al., 2013; Bokhorst et al., 2009;

Nawrocki et al., 1996). In studies from the Czech Republic, Poland and western Ukraine, the highest susceptibility values are generally found in the non-weathered loess layers while the lowest values are found in the most pedogenically developed horizons (Antoine et al., 2013;

Nawrocki et al., 1996; 1999). In the non-weathered loess and illuvial (accumulation) horizons, the main carriers of magnetic minerals are magnetite and maghemite (Nawrocki et al., 1996;

Maher & Thompson, 1991). In these layers, its less susceptible antiferromagnetic properties give hematite a significantly lower relative impact on susceptibility values compared to that of magnetite and maghemite. Hematite is however more stable than magnetite in a water environment where oxygen content is reduced and sulphur is available, as it commonly occurs as inclusions in acid-resistant minerals (e.g. quartz) while magnetite often occurs as

“unprotected” isolated grains (Evans & Heller, 2003; Nawrocki et al., 1996; Snowball, 1993).

These waterlogged conditions are characteristic of gley soils that periodically formed in the

European periglacial tundra environments during interstadials (Antoine et al., 2013; Nawrocki

et al., 1996). Gley soils therefore represent an environment in which highly susceptible

magnetite has been degraded and destroyed and the less susceptible hematite was preserved,

resulting in significantly reduced magnetic susceptibility compared to the loess (Antoine et al.,

2013; Nawrocki et al., 1996; 1999; Heller & Evans, 1995). Non-gleyed interglacial paleosols

have higher susceptibility values, as drier climate conditions allow for detrital magnetite to be

preserved and for pedogenic iron oxides to form and accumulate (Nawrocki et al., 1996). The

range of climatic conditions in which magnetic susceptibility is straightforwardly applicable

has therefore been questioned (Hosek et al., 2015; Bokhorst et al., 2009). An understanding of

site-specific conditions is crucial for a correct interpretation of how magnetic properties in loess

relate to paleoclimate.

11

Figure 4. Map of European loess deposits including last glacial maximum extent and distribution of periglacial environments. Red star marks the study site of Biały Kościół in SW Poland (Schaetzl et al., 2018).

Factors such as the mineralogy and source of the susceptibility carriers, and the effects of post- depositional alteration must be accounted for to reliably determine causal relationships and for correlation with other paleoclimate archives (Jackson et al., 1998; Maher, 1998).

2.2.2 Frequency dependent susceptibility

There are several other less well-known low field magnetic susceptibility parameters that can be measured and used to investigate sediment properties of grain size and mineralogy of magnetic minerals. In addition to composition and grain size, susceptibility measurements also depend on the frequency of the applied magnetic field (Hrouda, 2011; Jackson, 2003-2004:

Jackson et al., 1998; Dearing et al., 1996).

12

Low field magnetic susceptibility can be measured at different field frequencies (wavelengths (Hz)). Resulting mass bulk susceptibility measurements can vary depending on operating frequency (Evans & Heller, 2003). This parameter of frequency dependency (X

) is in rocks, soils and environmental materials traditionally thought to primarily reflect the interplay between SP-particles and SSD (stable single domain) or MD (multi-domain) magnetic particles (Hrouda, 2011). The difference in grain size produces a loss of susceptibility with increased operating frequency (Hrouda, 2011; Néel, 1949). Changing the operating frequency gives magnetic grains more or less time to react to the external AC-field. Increasing the frequency of measurements effectively shortens the time needed for magnetic alignment (i.e. magnetization), shifting the SP-SSD boundary to smaller grain sizes (Moskowitz, 1991). A narrower distribution of the highly-susceptible SP-particles produces a relative loss in the susceptibility signal at higher frequencies (Fig. 5). Frequency dependence is higher (lower) for a sample with a higher (lower) relative amount of fine-grained ferrimagnetic minerals and varies for different magnetic minerals, with magnetite generally having a higher frequency dependence than maghemite (Hrouda, 2011; Evans & Heller, 2003). Frequency dependence can therefore be used to infer properties of mineralogy and grain size of a sample that may be valuable for paleoclimatic studies (as a quick and easy alternative to more labour-intensive methods) and may also allow for a more reliable interpretation of bulk susceptibility data (Hrouda, 2011;

Jackson et al., 1998; Dearing et al., 1996).

Figure 5. Low field magnetic susceptibility plotted against particle volume of magnetite grains at room temperature. Highest susceptibility values are measured at the largest particle volume (before transition to SSD- state) and the lowest operating frequency (Hrouda, 2011).

13

Frequency dependence in loess is generally higher in the paleosols than in the loess, indicating a higher relative content of SP-particles in the SP-SD transition state (Evans & Heller, 2003).

This is thought to reflect the presence of secondary ultrafine ferrimagnetic minerals that, due to the small grain size range, are unlikely to include primary minerals and which therefore likely are the result of pedogenic processes (Dearing et al., 1996). This is supported by a positive correlation between magnetic susceptibility- and frequency dependence values (Hosek et al., 2015). A negative correlation (as observed in Germany, Ukraine and Poland) may be caused by depletion of primary magnetic minerals (SD- and MD-particles) during waterlogged conditions (gleyification)(Hosek et al., 2015; Taylor et al., 2014; Baumgart et al., 2014). This would lead to a shift toward a higher relative content of SP-particles, causing susceptibility values to decrease as frequency dependence values increase (Hosek et al., 2015).

2.2.3 Out-of-phase susceptibility

Low field magnetic susceptibility measurements are composed of one component that varies in phase with the applied magnetic AC-field and one component that does not (Hrouda et al., 2013; Jackson, 2003-2004; Néel, 1949). For the in-phase susceptibility parameter, this means that the relationship between the applied field and the induced magnetization does not vary with time and maintains a constant ratio value (Jackson, 2003-2004). In-phase susceptibility is used to measure the previously mentioned frequency dependent parameter and is as such relatively widely used for various geological and environmental purposes (Hrouda et al., 2011; 2013).

The out-of-phase susceptibility parameter (X

) has not received much recognition although it is routinely measured automatically with the multi-function Kappabridge, along with the in- phase susceptibility component, and may allow for interpretations of magnetic granulometry (Hrouda et al., 2013).

For susceptibility measurements at low frequencies and/or for a higher proportion of SP- particles, sample magnetization responds in lock-step to the alternating applied magnetic field.

As frequencies and/or the proportion of SSD-particles increase, there may occur a phase shift (denoted , also known as lag, phase angle or simply phase) as magnetization increasingly lags behind the applied field (Fig.6)(Hrouda et al., 2013; Jackson, 2003-2004; Dearing et al., 1996).

If there is a sufficiently wide distribution of grain sizes in the SP-SSD transition state, an

increased frequency will increasingly “push” SP-grains across to the stable SD-state by moving

the boundary to smaller grain sizes (Hrouda et al., 2013).

14

Figure 6. The variation of in-phase (X’) and out-of-phase susceptibility (X’’) with magnetic particle volume for different operating frequencies (1, 4, 16 kHz). The in-phase component drops sharply after reaching maximum magnetization values at the transition from SP- to SSD-states, while the out-of-phase component shows close to non-existent values at smaller and larger grain sizes, with a bell-curve shape in between. Both components show the highest values for the lowest operating frequency, decreasing with increasing frequency until reaching the smallest values at the highest operating frequency (Hrouda et al., 2013).

The time-varying component of the applied field is thereby shortened while the time required for particles to become magnetized increases. This viscous relaxation causes magnetization to lag behind the time-component (frequency) of the applied field (Jackson, 2003-2004).

Some loess-paleosol sequences have demonstrated a close linkage between the out-of-phase

signal and the frequency dependence of the in-phase signal (Hrouda et al., 2013). New

parameters have therefore been developed (based on the concept of relaxation time by Néel,

1949) that approximately converts the out-of-phase susceptibility to frequency dependent

susceptibility. The two parameters can then be used in combination to infer properties of

magnetic granulometry (SP-particles) that may prove useful in environmental studies (Hrouda

et al., 2013). Interpretations must however take into account the potential influence of other

magnetic fractions on the phase signal. Low phase values may reflect the presence of

paramagnetic and/or MD ferromagnetic (coarse-grained) fractions as these tend to decrease the

phase signal. Conversely, high phase values may indicate the influence of the diamagnetic

fraction, as it tends to increase the phase, rather than reflect high amounts of SP-particles

(Hrouda et al., 2013).

15 2.3 Dust provenance analysis

A fundamental aspect of dust research is identifying dust source areas. Knowledge of dust origin is the critical first step towards understanding mechanisms of dust production and emission, and processes of dust entrainment, transport and deposition (Muhs et al., 2014). As these factors are controlled by, and exert significant influence on, global climate, provenance studies of loess are an integral part of paleoclimatic research (Choobari et al., 2014; Maher et al., 2010; Harrison et al., 2001).

Dust provenance studies are commonly undertaken using methods of geochemical fingerprinting, i.e. attempting to match the mineralogical-, isotopic- or elemental composition of loess deposits with that of potential source areas (Muhs et al., 2014; Harrison et al., 2001).

This is however not a straight-forward task, as dust from different source areas mix and potentially homogenize in the atmosphere during transport, complicating an identification of multiple individual sources (Muhs et al., 2014; Liang et al., 2013; Buggle et al., 2011; Pye, 1995). The effects of weathering, grain-size sorting, recycling and admixing of non-aeolian sediment may alter loess composition to such a degree as to no longer allow for a reliable distinction between possible parent rocks (Kemp, 2001). Successful provenance techniques therefore either account for these effects or remain unaffected by them (Muhs et al., 2014).

2.3.1 Single-grain zircon U-Pb dating

Absolute dating of mineral ages in a sample is arguably the most commonly applied approach in provenance studies of sediments, as these can be used to identify source regions by correlation with protolith (parent rock) ages or with measured assemblages in sedimentary sources (Pullen et al., 2014).

U-Pb dating is a radiometric technique which determines the age of a mineral by measuring the ratios of radioactive parent (

U,

U) to daughter (

Pb,

Pb) isotopes. Uranium decays to lead (

U-

Pb,

U-

Pb) via a series of intermediate daughters which have known half- lives (decay rates) of tens to hundreds of thousands of years, making the technique applicable on a wide variety of timescales (White, 2015).

U-Pb dating of detrital zircons (ZrSiO

; Fig. 7) is by far the most robust and widely used

geochronological method in sedimentary provenance studies (Pullen et al., 2014; Fedo et al.,

2003). The development of increasingly fast and cost-effective analytical laboratory

instruments, such as Sensitive High-resolution Ion Microprobe (SHRIMP) and Laser Ablation

Inductively Coupled Mass Spectrometry (LA-ICP-MS), allows for large datasets relevant for

provenance studies to be analysed (Vermeesch, 2012; Thomas, 2011; Fedo et al., 2003).

16

Figure 7. Cathodoluminescence (CL) images of extracted zircons from the Biały Kościół loess samples (imaged at the Arizona Laserchron Center), displaying differences in grain morphology.

Zircons are heavy minerals (4.65 g cm

) which are extremely resistant to chemical- and physical weathering and therefore act as “closed systems” (i.e. retain isotopic composition) throughout most Earth surface and crustal processes and possibly even during short excursion into the mantle (Hawkesworth & Kemp, 2006; Pullen et al., 2014). Zircon grains are formed in silica-rich magmas and commonly occur as accessory minerals in most intermediate to felsic igneous rocks and are abundant in siliclastic sediments (Vermeesch, 2012; Fedo et al., 2003).

The high concentrations of U (100-1000 g g

), the exclusion of Pb during crystallization (ng n

) and the retainment of U and Pb at geologically high temperatures (Tc ≈ 900C) make zircons ideal provenance indicators, as calculated U-Pb ages are likely to reflect the crystallization ages of source rocks (Thomas, 2011; Pullen et al., 2014). The varied external morphology and complex internal zonation patterns of zircon grains can be used as indicators of source terranes (by correlation of morphologic features), recycling (euhedral/

subhedral/rounded grains) and metamorphic events subsequent to crystallization (overgrowth zones/rims)(Meinhold et al., 2008; Pupin, 1980). Zircons therefore contain information of the geological history of their proto-rocks (Vermeesch, 2012).

Provenance analysis of zircons is based on correlating U-Pb age population distributions of samples with that of potential proto-rocks, rather than focusing on absolute formation ages (Thomas, 2011; Fedo et al., 2003). Sediment deposits likely include multiple source components as: i) multiple magmatic events (crystallization ages) may have occurred in any one location; ii) tectonic forces may result in crustal rocks of different ages being in close proximity; iii) older sediments may be recycled and mixed with younger “fresh” sediments and;

iv) multiple source regions may drain into the same locality (Thomas, 2011).

17

The analysed sample should therefore ideally include evidence of all provenances so as not to risk making inaccurate geological interpretations (Pullen et al., 2014; Fedo et al., 2003).

Complexity is added by the effect of differences in rock properties, grain size and surface processes on the resulting detrital zircon abundance from a specific source: i) rocks with low zircon content will be underrepresented; ii) coarser, heavier grains will behave differently during transport compared to finer, lighter grains; and iii) weathering and erosion during transport and after deposition may preferentially degrade older grains with higher U-content (Fedo et al., 2003). Furthermore, as provenance analysis of detrital zircons assumes the random distribution of age fractions in a sample, artificial bias that may result in non-representative results may be introduced throughout the different steps of sampling, laboratory mineral separation and preparation, and age generation (Pullen et al., 2014; Fedo et al., 2003)

The number of age fractions present in a sample are likely to be unknown before analysis.

The minimum number of analysed grains required to adequately reflect the entire detrital zircon age population is therefore of fundamental importance and has been the focus of much debate (Pullen et al., 2014; Vermeesch, 2004). The probability of identifying all age fractions in a sample aliquot (subsample) generally increases with the number of grains analysed (n). For results to be geologically meaningful, all age fractions in a sample should be detected more than once (Pullen et al., 2004). In earlier provenance studies, it was suggested that 117 dates were required to be 95% certain that no age fraction greater than 1 in 20 would be missed (Vermeesch, 2004). This would however limit inter-sample comparisons to solely focusing on the presence or absence of age fractions, as opposed to the increasingly large sample sizes (n=300-1000) used in recent studies that may allow for quantitative comparisons between relative proportions of detrital sample ages (Pullen et al., 2014).

Another contentious issue is how to best visualize and compare detrital age data. Concordia diagrams are used to display U-Pb isotopic data and relevant analytical information (e.g. sample size, accuracy, precision) (Fedo et al., 2003). These diagrams are however not ideal when dealing with large-n datasets that tend to produce overlap between sample points and reduce readability. Histograms, Probability density distribution plots (PDP) and, more recently, the Kernel Density estimator (KDE) are therefore more commonly used to visualize detrital zircon data, often in combination (Pullen et al., 2014; Vermeesch 2004; Fedo et al., 2003).

As the use of large-n datasets have increased, the statistical validity of PDPs has come into

question and the Kernel Density estimator (KDE) has been suggested to be a more robust

statistical tool (Vermeesch, 2012; Pullen et al., 2014).

18

These diagrams both show the relative likelihood of the age population of a sample but while the standard deviation in the PDP is determined by the analytical precision, the standard deviation in the KDE is determined by the local probability density, reducing the risk of

“oversmoothing” (underestimating precision in age peaks) as analytical precision decreases in the older age fraction (Fig. 8)(Vermeesch, 2012).

Detrital zircon age populations are often by themselves not sufficiently unique to reliably identify sediment provenance. Interpretations of likely source rocks must therefore fit within a coherent framework of sediment dispersal paths and -recycling, stratigraphy, sedimentology, tectonics and paleogeography, and it is recommended to use complementary analytical methods to build a more complete geological picture (Thomas, 2011).

Figure 8. Illustration of the two main problems with PDPs. Top row shows how PDPs (black curves) break down when analytical precision (data quality, n=117) is high (c), compared to the synthetic “true age” distribution (Probability Density Function (PDF), green) obtained by measuring an infinite number of randomly selected detrital zircons with infinite precision (a) and the KDE (b). Bottom row illustrates the bias of PDPs when sample size is high (n=10 000) and analytical precision is low. The second mode in the smoothed version of the PDF (d) has been more smoothed as analytical errors increase with increased ages. As sample size increases, the KDE (e) increasingly resembles the true distribution (e). Due to the inherent dependence of PDPs on analytical precision, the PDP oversmooths the second mode, introducing bias as analytical errors increase (f)(Vermeesch, 2012).

2.3.2 Dust sources, -emissions and paleoclimate

Both glaciogenic and non-glaciogenic dust sources have been proposed as important for the production and emission of European dust during (at least) the Weichselian (Moine et al., 2017;

Rousseau et al, 2014; Sima et al, 2009; Smalley, 1995).

19

Glaciogenic sources include the periglacial areas bordering the Northern Hemisphere ice sheets and ice caps and at the foothills of mountain glaciers (Fig.4), where significant amounts of coarse material were reduced to silt-sized mineral dust through glacial grinding, frost action and glacifluvial abrasion (Rousseau et al., 2014; Evans & Heller, 2003; Smalley, 1995). The produced material was then transported from source area to the place of deposition by the wind, either directly or after initial transport by fluvial processes (Haase et al., 2007; Evans & Heller, 2003). The wide braided periglacial European river systems (e.g. Rhine, Danube, Dnieper, Rhône) were important conveyors of sediment during warmer melting periods, and source areas for dust emissions during colder, arid periods when reduced river flow exposed former river beds and surrounding plains to aeolian reworking (Rousseau et al., 2014; 2018; Újvári et al., 2012; Buggle et al., 2008).

Non-glaciogenic dust sources include the areas that experienced strong aeolian deflation during windy and arid glacials and stadials, due to reduced soil moisture, reduced vegetation cover and/or as the growth of continental-sized ice sheets caused sea levels to drop and expose more of the continental shelf to aeolian erosion (i.e. large parts of the present North Sea and English Channel)(Fig. 9)(Moine et al., 2017; Rousseau et al., 2014; Sima et al., 2009). Non- glaciogenic sources may also include areas of high topography which produce dust by physical weathering of exposed rock (tectonically) such as periglacial environments with cold climate processes, rather than by glacial processes, which is then directly or indirectly deposited by aeolian processes (Újvári et al., 2012; 2014).

Dust modelling results indicate that the main Middle and Upper Pleniglacial (i.e. ~ marine oxygen stage (MIS) 3 and MIS 2) sources of the thick loess deposits that make up Europe’s

“northern loess belt” (located between 48N and 52N) are located in the same latitudinal band (Rousseau et al., 2014). Air masses transporting dust are suggested to have moved west-to-east as the presence of large Northern Hemisphere ice sheets and ice caps created a “wind corridor”

across the continent (Fig. 9)(Rousseau et al., 2014; Sima et al, 2009).

Loess deposits across Europe from west to east, demonstrate temporal variability in grain size, sedimentation rates and soil development (Rousseau et al., 2011; Antoine et al., 2009).

Periods of development of soils, tundra gley horizons and periglacial structures are associated

with North Atlantic warm phases, when dust emissions were reduced by higher temperatures

and precipitation, higher sea levels and increased protective vegetation cover. Periods of high

sedimentation rates and coarser grain sizes are associated with North Atlantic cold phases, when

dust emissions were enhanced by lower temperatures and precipitation, lower sea levels and

reduced vegetation cover (Rousseau et al., 2014; Sima et al., 2009; Antoine et al., 2009).

20

Figure 9. Map showing European loess deposits, Last Glacial Maximum ice sheets and ice caps and the exposed continental shelf in large parts of the present North Sea and the English Channel. Proposed primary dust sources marked in bold text: Fennoscandia, the Eastern Alps, the Bohemian Massif (BM), Moravia (MO), the Sudeten mountains (SU) and the Carpathians. Red star marks the study site of Biały Kościół in SW Poland (Modified from Rousseau et al., 2018).

North Atlantic warm phases correspond to warm events and abrupt temperature increases recorded in Greenland (referred to as Greenland Interstadials (GIS) and Dansgaard-Oeschger (D/O) Events respectively). Conversely, North Atlantic cold phases correlate with Greenland cold events (Greenland Stadials (GS)) and periods of intense ice-berg rafting (Heinrich Events)(Rousseau et al., 2007; Antoine et al., 2001; Johnsen et al., 2001). High-resolution radiocarbon age models have demonstrated synchronous Middle and Upper Pleniglacial variations in glacial dust deposition in Greenland and eastcentral Europe on centennial- millennial timescales and have correlated Greenland interstadials with specific soil horizons in the west European loess (Moine et al., 2017; Újvári et al., 2017). This cyclical depositional pattern and the high-frequency variations in the west-to-east European loess belt are therefore suggested to have been caused by the influence of Greenland stadials and -interstadials and Heinrich- and Dansgard-Oeschger Events on North Atlantic air masses, Fennoscandian ice sheet dynamics, sea level changes and surface conditions (Újvári et al., 2017; Moine et al., 2017; Rousseau et al., 2007; 2011; 2014; Sima et al., 2009; Antoine et al., 2001). This highlights the importance of European loess archives, not solely for reconstructing the paleoclimatic evolution of Europe, but for a more complete and correct understanding of the interplay

Fennoscandia

E. Alps

BM MO

SU

21

between global climate factors and changes in dust source areas and dust emissions (Rousseau et al., 2018; Újvári et al., 2017).

2.3.3 Primary and secondary dust sources

Primary sources are the protoliths (parent rocks) from which dust has been produced by physiochemical weathering processes (Fedo et al., 2003; Kemp, 2001). Secondary sources refer to sediments in which material from primary sources was deposited and/or transported before subsequent aeolian reworking and final deposition (Haase et al., 2007). As an accurate identification of dust provenance is based on reconstructing dust pathways from source area to place of deposition, one must consider and account for the potential influence of both primary- and secondary dust sources.

A combination of trace element- and isotopic data and model results indicate that last-glacial dust sources for loess deposits in west and east (Serbia, Ukraine) Europe and east Germany appear to be clearly distinct and mainly influenced by proximal sources, with transport primarily occurring by westerly storm events at low elevations over local to regional distances (tens to tens of thousands of kilometers)(Rousseau et al., 2014; Antoine et al, 2009). The exposed continental shelf in the present North Sea and English Channel during the low sea level stands of the last glacial is highlighted as a major possible dust source to primarily western Europe, but possibly also further inland along the northern loess belt (Fig. 9)(Rousseau et al., 2014; Antoine et al, 2009).

Detrital zircon U-Pb age data and heavy mineral assemblages of Pleistocene Rhine River terrace sand indicate similar source assemblages along the river. The main U-Pb age peaks reflect primary sources related to the Variscan orogeny (300-400 Ma), Baltica or the mid- German crystalline rise (400-500 Ma), and Cadomian terranes (ca. 570 Ma), rather than the Alpine orogeny (Krippner & Bahlburg, 2013). The Variscan orogeny occurred when the Euramerican (Laurissian) continent collided with Gondwana to form the supercontinent of Pangea in the late Paleozoic, while the Cadomian orogeny occurred on the margin of the continent of Gondwana in the late Neoproterozoic (550-650 Ma). Variscan, Baltic (Caledonian) and Cadomian terranes are found across Europe, including in south Germany, Bohemia and southern Poland (Krippner & Bahlburg, 2013).

Trace element- and isotopic composition data and zircon and rutile U-Pb ages suggest that

last-glacial Austrian loess is derived from proximal dust sources in the Bohemian Massif and

the Eastern Alps (depending on sample location).

22

Primary sediment transport is indicated to be strongly influenced by fluvial networks (rather than wind direction) and aeolian processes are suggested to be a second mode of transport (Újvári et al., 2013; 2015).

Major- and trace element, mineralogical-, isotopic- and zircon U-Pb data from the Hungarian Danube loess also suggest the importance of proximal sources for the last 0.8 Ma, pointing to the Alps, the Carpathians and local highlands as primary dust sources and local alluvial fans and floodplain sediments as secondary sources (Fig. 9)(Újvári et al., 2008; 2012; 2014).

Major- and trace element composition, element ratios and geomorphology of southeast European loess indicate that i) the Carpathians and the Austrian Alps are likely primary sources of the Serbian Danube loess, drained by the Danube river and its tributaries (secondary sources) before aeolian deflation of alluvial sediments; ii) glaciofluvial sediments from the Fennoscandian ice sheets (primary source) in Ukraine and Russia are likely secondary sources of the Ukrainian Dnieper loess; and iii) that the Romanian loess is likely derived primarily from Danube alluvial deposits (secondary source) and from local Ukrainian glaciofluvial sediments (secondary sources) from the Fennoscandian ice sheet (primary source)(Buggle et al., 2008).

A recent study using U-Pb geochronology of detrital zircons in modern sands from the lower Danube and some of its largest tributaries, suggest primary sources to be the Romanian Carpathians and their foreland, with smaller contributions from the Balkan Mountains and from the east European Craton. Results indicate that zircon U-Pb ages primarily reflect igneous crystallization ages during the late Proterozoic, with mostly Cambrian and Ordovician sources (ca 600-440 Ma). A prominent magmatic Variscan peak at 320-350 Ma present in all samples was however left unexplained, as the Variscan source terranes that can be found in the nearby Carpathians were deemed too small to account for the relatively large production of zircon grains (Ducea et al., 2018).

Magnetic susceptibility data and detrital zircon U-Pb data suggest that loess in southern Ukraine was transported by the Dniester River from dust source areas in the Carpathians and the adjacent Podolian highlands (reflecting Neoproterozoic and Paleozoic ages) before subsequent aeolian deflation. Reconstructions of wind orientation indicate that northerly winds dominated during glacials and northwesterly winds during interglacials (Nawrocki et al., 2018).

Due to the lack of provenance studies in general, and single-grain studies in particular,

interpretations of Polish dust sources appear to be primarily based on sedimentological and

geomorphological interpretations of loess deposits, and on micromorphological features of

sediment grains. Smalley and Leach (1978) conclude that the geographical distribution and

sediment characteristics (e.g. grain size, mineralogy, structures) of central European loess

23

deposits indicate an obvious glaciogenic origin for loess in central Europe and the Carpathians, produced by aeolian deflation of periglacial deserts (at the margin of the Weichselian ice sheet).

Loess deposits in west, southeast and east Europe were concluded to have been formed by reworked silty material from the major European river systems (e.g. Rhine, Dnieper). However, as stages of loess accumulation occurred even as the Weichselian ice sheet had retreated from the central European lowlands, this hypothesis was revised by Badura et al., (2013). Based on late Pleistocene sedimentary analysis and landforms in Lusatia and western Silesia (southwest Poland), the silty materials that form the central European (including Polish) loess deposits are proposed to have been supplied by major paleoriver networks (in addition to periglacial deserts) draining source areas in the Carpathians, Sudeten Mountains and central German highlands.

The retreat of the Weichselian ice sheet is suggested to have re-routed major river systems and produced a paleoriver valley (the Great Odra Valley) during the late Pleniglacial, which was then buried by weathered sediments from surrounding highlands. These valley sediments were subsequently reworked and re-deposited in what is now the northern loess belt by the prevailing northwesterly winds, and in the Carpathian forelands by westerly winds (Badura et al., 2013).

Pleistocene loess deposits from south Poland have also been studied on a micromorphological level by Mroczek (2013). Microfeatures related to lithology, pedogenesis and cryogenic processes found in the primary (supposedly unweathered) loess was interpreted as evidence for older pedogenic modification at a previous location, before re-deposition and re-working at the second location as younger loesses were formed. This in turn was suggested to indicate a direct link between the source of primary Weichselian loess (L1) and buried interglacial soils (S1), and the possibility of very proximal dust sources for the Polish loess.

In summary, previous provenance studies point to the eastern Alps, the Carpathians,

Moravia, the Sudeten Mountains, the Bohemian Massif and Fennoscandia as the main primary

sources for central European dust (Fig. 9), while the fluvial- and glaciofluvial deposits of the

major European (paleo-) river systems, in addition to the exposed continental shelf of the

Present North Sea and English Channel, are indicated as important secondary sources.

24

3 Materials and methods

3.1 Study site and sampling

The Biały Kościół (N 5043’38.3’’, E 1701’29.5’’, elevation ~180 m.a.s.l.) loess section is located in an old clay pit on the western slopes of the Oława river in the forelands of the Sudeten Mountains in southwest Poland (Moska et al., 2011)(Fig.10). The study area has a mean temperature of -1.5C during winter and 17.5C during summer, and a mean annual rate of precipitation of 600mm (Komar et al., 2009). The loess section is ~9 m thick and consists of five main stratigraphic units: i) three polygenetic soil complexes (Holocene top soil (S0), last glacial interstadial soil (L1SS1), last interglacial soil (S1)) and ii) two calcareous loess units (last glacial younger/middle- and lower loess (L1LL1, L1LL2)(Fig 11). The abbreviations for the loess and paleosol units are in accordance with the widely accepted Chinese “L-S”-system, where “L” represents loess and “S” represents soils (paleosols)(Markovic et al., 2015).

The cleaned loess profile was sampled for low-field magnetic susceptibility (MS) measurements and single-grain detrital zircon U-Pb geochronology.

Figure 10. Map of loess distribution in southwest Poland. Red star marks the study site of Biały Kościół (Modified from Moska et al., 2011).

25

Figure 11. Illustration of stratigraphic-, geochronological- and sample information from the Biały Kościół loess profile. Circles illustrate sample points (red= U-Pb ages, blue = MS-measurements). Horizontal red lines separate stratigraphic units (S0=Holocene soil, L1LL1=younger/middle loess, L1SS1=interstadial loess, L1LL2=lower loess, S1=top interglacial soil/interstadial soil). Included are Optically Stimulated Luminescence (OSL) ages (ka) with errors from Moska et al. (2011) and stratigraphic visualization and description from Jary et al. (2016).

Thirty-seven MS-samples were collected at a resolution of 20-40 cm throughout the profile and a total of five samples were collected for U-Pb dating from each of the five stratigraphic units S0 (0.5m), L1LL1 (2.8m), L1SS1 (6.1m), L1LL2 (7.35m) and S1 (8.7m)(Fig. 11).

The results of this study will be placed into the detailed stratigraphic and geochronological context provided by previous publications of the Biały Kościół loess section (Fig.

11)(Krawczyk et al., 2017; Jary et al., 2016; Jary & Ciszek, 2013; Moska et al. 2011; 2012).

These previous publications also include data of magnetic susceptibility, grain size distribution,

calcium carbonate content, organic carbon content and micromorphological features that may

prove relevant and useful to compare with the results and interpretations of this study, to build

a more complete picture of dust provenance and paleoclimatic evolution.

26 3.2 Magnetic susceptibility

Samples were oven-dried at ~40C until completely dry and carefully mortared to remove aggregates, while at the same time avoiding to break grains, before being weighed and placed into 7 cm

-sized plastic cubes until full (Fig. 12). Magnetic susceptibility was measured at room temperature using the Agico MFK1-FA (Multi-Function Kappabridge;

https://www.agico.com/) at the Uppsala University, Department of Earth Sciences (Fig. 12).

Measurements were performed with an AC (alternating current) magnetic field amplitude of 200 A/m at three different frequencies, i) 976 Hz (low frequency, F1); ii) 3904 Hz (intermediate frequency, F2); and iii) 15 616 Hz (high frequency, F3), to allow for a determination of frequency dependence. Before beginning measurements of actual samples, the Kappabridge was calibrated with a standard of known susceptibility (from Agico), calibration was performed at each frequency. To increase understanding of measurement uncertainty, samples were measured three times (each measurement taking ca. 20 seconds) at each of the three frequencies (a total of 9 measurements per sample) and the reported magnetic susceptibility is the arithmetic mean of triple measurements (Dearing et al., 1996). Instrument sensitivity for the measurement is 2x 10

(S1) and consequently the calculated χ is in the order of 10

m

kg

(Hrouda, 2011).

Measurements in this study are typically at least 3-4 orders of magnitude higher than the instrument sensitivity. Instrument accuracy for absolute calibration is ± 3%.

Figure 12. Left: Photo of prepared sample cubes. Right: Photo of the MFK1-FA Kappabridge at the Uppsala University, Department of Earth Sciences (Sweden).

27

The applied magnetic field (H) and the induced magnetization (M) are measured simultaneously and the data acquisition software SAFYR (AGICO) automatically generates the in-phase and out-of-phase magnetic susceptibility components data, as well as the change in angle between two phases.

The volume-dependent bulk susceptibility  (induced magnetization per unit field) is a dimensionless parameter, defined as the ratio between M (A/m) and H (A/m)(Evans & Heller, 2003):

 = M/H [dimensionless] (1)

To obtain the mass-specific magnetic susceptibility (χ),  is divided by sample density ()(Evans & Heller, 2003):

χ = / [m

/kg] (2)

The sample size used in this study is smaller (7 cm

) than the standardized default volume (10 cm

) used by the software to automatically generate χ.  This means that absolute χ will be somewhat underestimated, but by the same relative amount for each sample.

Frequency dependent magnetic susceptibility (X

) is commonly expressed as a percentage of loss in susceptibility between low (X

) and high (X

) field frequencies (Hrouda, 2011):

X

= 100(X

– X

)/ X

[%] (3)

The frequency dependent parameter (f

) is however best described by the out-of-phase component of susceptibility (characteristic of the SP-SSD transition state) which has a logarithmic relationship with the operating frequency. It has therefore been suggested to use the normalized frequency dependent parameter (X

) expressed as (Hrouda, 2011):

X

= X

/(ln f

– ln f

) [%] (4)

In this study, calculations of X

are modified after the re-written equation presented in Hrouda

(2011), which generates positive values instead of negative, to increase comparability and

readability of the plotted data.