Loess and other Quaternary sediments in Germany

ABSTRACT Geo- and palaeoecological studies focusing on the late Pleistocene require a detailed knowledge of the spatial distribution of aeolian sediments. In Germany, existing maps are either on large scales, have a regional focus or show significant inaccuracies such as artificial boundaries within different geological units. To obtain a more detailed, seamless map of the distribution of aeolian sediments and their potential source areas, we combined and reanalysed available geodata, using a Geographical Information System. The resultant maps (scale: approx. 1:2,600,000) show the link between source areas and the late Quaternary aeolian deposits in Germany and can provide one context for further work on, e.g. palaeogeographical studies. This work was compared with other already published datasets and the problems of sediment mapping at a small scale were discussed.

'Loess maps' have a long history in Quaternary science. Grahmann (1932) developed the first map of loess distribution in central Europe, which was adapted and compiled with source areas by Flint (1971). Fink, Haase, and Ruske (1977) created a map for Western, Central and Eastern Europe at a scale of 1:2,500,000 in collaboration with researchers from various countries. This map was completed and digitised by Haase et al. (2007). Other maps published in scientific studies during the last decades are characterised by either a regional (Antoine et al., 2003;Lill & Smalley, 1978) or a global scale (Muhs & Bettis, 2003;Pécsi, 1990).
Maps at a small scale often lack accuracy or show inconsistencies such as artificial breaks at administrative boundaries. This is related to different terminological and methodological foci of the contributors (Haase et al., 2007). Most researchers accept these problems, so that the map according to Haase et al. (2007) is widely accepted in Quaternary science. Recently, however, there were some attempts to create more precise maps of loess distribution. For instance, Lindner, Lehmkuhl, and Zeeden (2017) used paedological information to fill gaps in the Haase map in western Romania. Bertran, Liard, Sitzia, and Tissoux (2016) used topsoil textural data from the Land Use and Cover Area frame Statistical survey database (LUCAS, Orgiazzi, Ballabio, Panagos, Jones, & Fernández-Ugalde, 2018;Tóth, Jones, & Montanarella, 2013) to create a map of aeolian sediments in Europe, with a focus on France.
To follow-up on the recent research, both methodologically and geographically, the aim of this study is the creation of a detailed, seamless map of the distribution of aeolian sediments and their potential source areas in Germany. In further studies, these maps could be used for palaeolandscape modelling. For these objectives, we reanalysed and combined free, online available data of the Federal Institute for Geoscience and Natural Resources of Germany (Bundesanstalt für Geowissenschaften und Rohstoffe, BGR), to get a detailed overview of the distribution of aeolian sediments within Germany.

Study area
Germany is located in Central Europe and according to Liedtke and Marcinek (2002) it can be divided into five main parts: (a) the north German lowlands, which are dominated by a low relief formed by Pleistocene glacial and periglacial processes; (b) the central part with mountains composed of folded Palaeozoic rocks and partly covering sedimentary bedrock; (c) the south German scarplands, which are dominated by cuestas of sedimentary bedrock; (d) the Tertiary molasse sediment basin including remnants of piedmont glaciations south of the Danube River and in the foreland of the Alps; and (e) the Alps with their high mountain environment in the south.
Regarding its geography, the loess distribution in Germany follows a specific pattern, which allows the division in four main loess regions (cf. Lehmkuhl, Zens, Krauß, Schulte, & Kels, 2016) (Figure 1): (1) The northern Central European loess belt between the former ice marginal position of the Weichselian ice sheet in the north and the German uplands (e.g. Rhenish shield, Harz Mountains and Ore Figure 1. Study area, modified according to Lehmkuhl et al. (2016). The distribution of aeolian sediments is according to Haase et al. (2007). The numbers refer to the main loess regions and the letters to the subregions listed in the text. Ice extent modified according to Ehlers, Eissmann, Lippstreu, Stephan, and Wansa (2004)

Methods and data sources
3.1. Source maps, spatial data and processing The presented maps build upon freely available data from the BGR. For the mapping of the potential source areas of aeolian deposits, the digital geological map of Germany with a scale of 1:1,000,000 was used (Toloczyki, Trurnit, Voges, Wittekind, & Zitzmann, 2006). Even though a dataset with a larger scale would be more precise, the nomenclature of those maps was too differentiated to be integrated into the present map. Therefore, we used this map, in order of generalisation. Generally considered dust source areas for loess deposits are the shelves of the North Sea and the English Channel (Antoine et al., 2013) and late Pleistocene accumulations of fluvial origin, including the areas covered by Holocene fills today that were part of the Pleistocene braided river systems (Smalley et al., 2009). In addition, the last glacial outwash-plains and fluvio-glacial valleys (German: Urstromtäler) are considered as dust source areas (Bertran et al., 2016;Smalley et al., 2009). These different units were extracted from the attribute table of the geological map (Toloczyki et al., 2006) and are shown in map A (Main Map). The shapefile for the dry shelves is modified according to Willmes (2015).
For the mapping of aeolian deposits, the digital geological map of Germany with a scale of 1:200,000 served as a major source (BGR, 2007). It is more precise than the 1:1,000,000 map and the loess distribution map of Haase et al. (2007), see Figure 1. This dataset contains 55 shapefiles corresponding to the 55 map sheets of the original analogue maps ( Figure 2). These shapefiles cover the area of Germany as well as adjacent regions. As the first step, the entire Quaternary sediments were exported from the attribute table of each shapefile. In the second step, the aeolian sediments were extracted. Hereinafter, those were subdivided into three major classes: loess and loess derivates, sandy loess, and dunes and aeolian sand (for a more detailed description of the classification, see Section 3.2 and Table 1). The resulting shapefiles were merged to obtain coherent datasets.
Generalisation of the data, especially for the detailed loess map, was conducted by using a self-built model ( Figure 3). The model performs an automated aggregation, simplification and smoothing to adjust the obtained data to the smaller scale of the resultant maps. The shapefiles were aggregated three times, whereby different values for the minimum hole size and the minimum area were applied. Therefore, the aggregation algorithm scans the shapefile in order to detect all polygons and holes in the shapefile (ESRI, 2016a). Shapefiles smaller than 1 km 2 were deleted and holes smaller than 10 km 2 were filled. Afterwards, the aggregated shapefile was simplified using a Bend-Simplify Algorithm and smoothed with a PAEK Algorithm (ESRI, 2016b(ESRI, , 2016cRegnauld & McMaster, 2007). The specific parameters for all generalisation steps are shown in Figure 3. For both maps, the same generalisation steps were chosen to create a smooth and consistent map design. Since the focus of the study is on Germany, the map contents were clipped with the German national borders, albeit the distribution of the sediments is cross-border.
To allow better orientation, major rivers (JRC, 2007;Vogt et al., 2007), administrative borders (gadm.org, 2018) as well as major cities were inserted in the map. Additionally, the 1-arc-second Japan Aerospace Exploration Agency (JAXA) Advanced Land Observing Satellite (ALOS) averaged digital surface model (DSM) was used to extract topographical information (JAXA EORC, 2016). This globally available elevation model is based on the AW3D dataset, which has a spatial resolution of ∼5 m. The high-resolution data were derived from stereographic multi-temporal satellite imaginaries, which were stacked, mosaicked and validated with existing height references such as the data from the Ice, Cloud and land Elevation Satellite (ICESat) mission as well as SRTM version 2 data. This dataset was down-sampled to offer the elevation data free of charge. The resultant DSM has a resolution of ∼30 m and a height accuracy of less than 5 m and is downloadable in 1x1 tiles Takaku, Tadono, & Tsutsui, 2014;Takaku, Tadono, Tsutsui, & Ichikawa, 2016). These tiles were mosaicked to obtain seamless and accurate elevation data for the study area.
We assume that sedimentological conditions of loess distribution (or dust deposition) and sediment relocation are related to relief conditions (Lehmkuhl et al., 2016), so a topographical analysis of the aeolian sediments was conducted. The ALOS DSM was clipped with the generalised shapefiles to obtain elevation data for the sediment distribution. The results are shown in Section 4. The coordinate reference system of the source shapefiles was the Gauß-Krüger-based German Main Triangle Net (Deutsches Hauptdreiecksnetz (DHDN) Zone 3, EPSG: 31467). In order to create a dataset which can be used internationally, the shapefiles were transformed to the European Terrestrial Reference System (ETRS) 1989 Lamberts Equal Area projection (LAEA; EPSG: 3035), as recommended by the European Union (INSPIRE, 2014).

Differences in mapping and loess definitions
The definitions of and the clear distinction between loess, loess-like sediments and loess derivates have been disputed for several decades (Pécsi & Richter, 1996;Pye, 1995;Smalley et al., 2001;Sprafke & Obreht, 2016). These discrepancies hamper a simple mapping of said sediments. The problems occurring from cross-border mapping can be seen, e.g. in the loess distribution map of Haase et al. (2007), where in some cases administrative borders correspond to the  Dunes and aeolian sands Aeolian sand Dunes and aeolian sand boundaries of loess distribution. An approach to overcome these cross-border issues, e.g. using geological and paedological data is given in Lindner et al. (2017).
In Germany, the mapping of loess and loess derivates started at the end of the nineteenth century and was mainly conducted by the Prussian Geological Survey (Preußische Geologische Landesanstalt, Wagenbreth, 1999). In the course of these mappings, only sediments with a thickness of more than 2 m were taken into account. This led to gaps, especially concerning the loess distribution. Additional problems occur because of federalism; each of the 16 federal states in Germany maintains its own geological survey. Thematic maps encompassing several administrative entities tend to show artificial breaks, e.g. at administrative borders (cf. Nilson, Köthe, & Lehmkuhl, 2007, and references therein). The 16 geological surveys compiled the 55 maps, based on a multitude of source data. These surveys spread over a time span of more than 30 years and were conducted by a large number of workers. The maps were digitised and published online in 2007. In this step, the data were not homogenised, leading to the variety of aeolian sediment classes. In some cases, even two maps of the same geological survey can show distinctive differences regarding the mapping or nomenclature (Figure 4). In order to simplify and synthesise the different maps, several classes were combined (Table 1).
The nomenclature of the source deposits is even more diverse. For example, the Munich gravel plain (see Section 4.4) is designated as 'lower terrace' as well as 'Würmian fluvial sediments' in two adjacent map sheets of the same geological survey (Figure 4). To avoid false mapping results, these sediments were extracted from source data with a smaller scale, which was the geological map at 1:1,000,000 scale (see Section 3.1).
Only sediments of primary aeolian origin were included. Sediments which were significantly altered such as by weathering (German: Verwitterungsbildungen)  or redeposition (German: Umlagerungsbildungen) were not taken into account, even though their primary origin could be aeolian. For information about the spatial distribution of relocated or colluviated loess, see Bertran et al. (2016).
The resultant map of the aeolian sediments is compared to two existing maps: the map of Haase et al. (2007) as the widely accepted reference and the map of Bertran et al. (2016) as a recent work with an approach differing from ours.

Distribution of loess and aeolian sediments in Germany
The following section focuses on the presentation and discussion of our results. The distribution of aeolian sediments is described for the different regions in connection with the potential source areas.

Aeolian and fluvio-glacial sediments in northern Germany
Especially in map A we show the distribution of alluvial fill, the lower terrace, the fluvio-glacial valleys (Urstromtäler) and the Weichselian fluvio-glacial deposits. The latter comprise mainly sandar (glaciofluvial outlet plains) which correspond to the different extents of the late Pleistocene ice sheet. Some of these glacio-fluvial outlet plains are related to the retreat of the Weichselian ice sheet (Frankfurt and Pommerian stages) and are, therefore, located further to the northeast than the maximum ice marginal position. We added these selected late Pleistocene sediments, as they represent a major local dust source.

The Northern Central European loess belt including sandy loess
Map B (Main Map) shows the belt-shaped distribution of loess on the northern margin of the German uplands (LR1). This pattern is almost parallel to potential source areasespecially the glacio-fluvial deposits and the Urstromtälerin northeast Germany. Grain size decreases with increasing distance from the Fennoscandia Ice sheet, especially in the eastern part of Germany: aeolian sand and sandy loess (in some areas and maps named Flottsand in German) can be found in proximity to the source areas (e.g. east of Hamburg respectively south of Berlin), whereas loess and its derivates can be found in distal positions. The North Sea was a vast outwash plain during the LGM and act as additional major particle source. It is noticeable that there are aeolian sand covers beneath the ice sheet of the LGM. These sediments were probably deposited during the late Weichselian or even during the Holocene after ice retreat (Hilgers, Murray, Schlaak, & Radtke, 2001;Koster, 2005;Küster & Preusser, 2010). Sandy loess is only present in elevations lower than 200 m a.s.l. (Figure 5). Most loess in this region is distributed in front of the German mountain ranges in lower elevations. This topographic barrier limits loess distribution to elevations roughly between 190 and 390 m a.s.l. The sand belt north of the loess region as represented by Zeeberg (1998) and Bertran et al. (2016) shows that the loess is juxtaposed to sand. In our map, the coversands are poorly represented as the uppermost 2 m were not included in the geological mapping.

The Upper Rhine Valley including the Mainz basin
Principal accumulation areas of loess in the Upper Rhine Valley are the rift flanks of the Upper Rhine graben. It should be noted that the loess distribution on the west bank of the Rhine is more distinctive than its eastern counterpart. The deposits in the west extend across the graben, whereas the loess in the east can be found at the margin of the Black Forest. The northeastern Upper Rhine valley is dominated by aeolian cover sands, which merge into the broad sand covers in the area of Frankfurt am Main. The sand material originated from the Pleistocene braided Rhine River. Elevations, especially in the Mainz basin north of the valley, exceed those in northern Germany (see LR2 in Figure 5). Another peculiarity of this region is the direct vicinity of loess deposits and their potential source areas. Southeast of Frankfurt am Main, triangular-shaped loess deposits are interposed with late Pleistocene fluvial deposits of the Rhine.

The basins of the German uplands
The third mentioned loess region are the basins within the German uplands. Due to the topographic limitation of these basins, the distribution pattern of their deposits is rather fragmentary. Exceptions are the broad basins in western Franconia (3b) and the Kraichgau (3a), southeast respectively south of Frankfurt am Main. The loess distribution spreads over elevations of 250-600 m a.s.l. (LR3 in Figure 5), with some outliers at both the upper and lower ranges.

Loess along the Danube
The loess distribution related to the Danube stretches along the upper reaches of the river. They can be mainly found between the Danube River in the north and the Würmian ice margin in the south. Akin to the Upper Rhine Valley, the sediments can be found directly next to the fluvial source areas. As a peculiarity, the Munich gravel plaina Pleistocene fluvial sediment body with an area of ∼1300 km 2has to be mentioned. The topographical distribution of the Danube loess in Germany shows the highest elevations, ranging between 400 and 650 m a.s.l., with a maximum of almost 800 m (LR4 in Figure 5).

Discussion
The datasets used and the methods applied must be discussed critically. The discrepancies in loess mapping mentioned in Section 3.2 limit the possibility to differentiate between various loess facies. In our map, loess and its derivates, which are given in more detail in the geological maps, were combined to one class. There is no possibility to differentiate between loess and loess loam for example. Prima facie, this may appear as a loss of information. As there are various definitions of loess, loess derivates, and loess-like sediments, sharp boundaries between those facies, as they are often presented in maps, feign high accuracies, which cannot be given at this scale. Such detailed distinctions cannot be extracted from geological maps, but should be investigated in selective (field) case studies.
Thus, the resolution is strongly dependent on generalisation. The model used was created in an iterative procedure, in which the intermediate products were evaluated thoroughly. Table 2 shows that the general area of the mapped sediments increased. These enlargements can be easily explained by the aggregation of small polygons. These were found in hilly areas with a distinctive surface geology. In the original map, all geological structures were taken into account. This leads to the fragmented visual appearance that is not suitable for a map of such a small scale. Due to the focus of our map on the aeolian sediments, those expansions were accepted to obtain a coherent design. The expansion of the polygons cannot only be seen in the increased total area, but also in the growth of the averaged polygon size. The values are an order of magnitude higher than the polygons of the original map, albeit most of them are smaller than the ones in the map of Haase et al. (2007), indicating a higher resolution ( Table 2). The distribution of aeolian sediments between the maps seems comparable, although our map's generalisation is smoother and shows less loss  Table 2. Comparison of the polygon areas between the raw data from the geological maps, the resultant map and the datasets published by Haase et al. (2007) and Bertran et al. (2016). Note: Area refers to the total area of the sediments (for abbreviations, see Figure 5; CL: colluviated loess, SS: silty sand), mean refers to the averaged size of the polygons of each shapefile.
of information. This can be seen in smaller mean values of the size of polygons as well as in the overestimation of widespread loess regions in the Haase map (LR1 and LR4). The comparison to the map published by Bertran et al. (2016) shows significant similarities and some differences: The mean areas are much smaller than in our map, resulting from raster source data with a resolution of 500 × 500 m. The loess distribution in LR1, 2 and 3 match, albeit there are regional differences such as in Saxony and south of the Harz Mountains. LR1 shows a particular high concordance in pattern and area. Larger discrepancies can be seen in LR4, where our map shows widespread loess deposits, whereas the Bertran et al. (2016) map shows only fragmentary patches. Those differences can be explained by the mapping of relocated and colluviated loess in this map, leading to an underrepresentation of loess and loess derivates. Accordingly, a large proportion of LR4 loess are mapped as loess derivates by Haase et al. (2007) and as colluviated loess by Bertran et al. (2016). Although sandy loess was not considered, silty sand is mapped as a transitional facies between sand and loess. The aeolian sand in northern Germany in the map of Bertran et al. (2016) shows the most striking differences to ours: while our map shows fragmentary distribution of sand, Bertran et al. (2016) mapped wide areas, whereas other aeolian sand deposits, such as in the area of Frankfurt am Main, are not included. This may be either an overestimation by Bertran et al. (2016) or more likely an underestimation in our map, since only sand deposits with a specific thickness were mapped, which are not always reached in northern Germany. As mapped by Zeeberg (1998) and Bertran et al. (2016), the coversands form an almost continuous band juxtaposed to the loess, which better highlight the grain-size gradient in the aeolian deposits.

Conclusions
This study provides detailed maps of the distribution of aeolian sediments and their potential source areas in Germany. The use of freely available geological data simplified the data provision, but is accompanied by some impediments: . The focus of geological mapping usually does not lie on the Quaternary deposits, which leads to a fragmentation of the sediment distribution in maps. . This method does not allow any assertions about the thickness or detailed timing of deposition of the sediments. There is no nationwide information about the mapping procedure, so we cannot state whether, for example, the uppermost 2 m was included into the geological mapping. Therefore, the sediment thickness cannot be determined precisely.
. The geological mapping spanned a long period of time and includes a multitude of different mappers. Due to the various procedures and loess definitions, artificial boundaries occur not just at administrative borders, but also within the same map sheet.
Here, we tried to overcome these issues by creating a seamless, coherent dataset. In order to achieve that aim, we propose the following solutions: . The extraction and generalisation of the shapefiles place the focus on aeolian sediments. Especially the generalisation eliminates the fragmentary pattern of the shapefiles. However, this must be taken into account, when discussing the data. The comparison to existing datasets is crucial to validate our map: the comparison to the map of Haase et al. (2007) shows that our map experienced an appropriate degree of generalisation. This leads to a high accuracy combined with good readability on a small scale. Compared to the map of Bertran et al. (2016), we see similarities in loess distribution, whereas the aeolian sand deposits in Germany in this map require further validation. . The thickness of a sedimentary body is indeed important information. Across a broad study area like Germany, information about the thickness feign a degree of accuracy, which cannot be given on this scale. Therefore, this issue could not be solved by our study. . Even though the distinction between different loess facies is not possible within our map, it provides an overview of the distribution of aeolian sediments. However, a more detailed approach is not possible on such a scale and should be provided by regional (field) case studies.
Thus, the map presents the geographical distribution of aeolian sediments within regional clusters, which are characterised by different topographical situations. The distribution of loess shows an increasing elevation southwards and also sandy loess displays a characteristic elevation below 200 m. The map indicates that geomorphological features like mountain ranges, valleys and graben structures variously act as barriers or sinks for loess accumulation. These features and the distribution of the aeolian sediments as well as their potential source areas allow conclusions about the palaeogeographical conditions in Germany. Our work highlights the value of the compiled geodata, which can be accessed freely at the CRC806 database at http://dx.doi.org/10.5880/SFB806.39.

Software
Mapping, processing and statistical analysis were done using ESRI ArcGIS 10.4.1 in focus of reproducibility and the broad availability of this software. Statistics were analysed using R 3.4.1 (R Core Team, 2014) and Microsoft Excel 2013. Main graphics were created using R 3.4.1 or Adobe Illustrator.