Anthropogenic relief changes in a long‐lasting lignite mining area (‘Ville’, Germany) derived from historic maps and digital elevation models

Humans constitute one of the main geomorphological agents in modern times. As an example, post‐mining regions represent a typical landscape of the Anthropocene. Strong relief modifications are particularly obvious with open pit mining. However, many existing mining areas are lacking detailed pre‐mining information for the quantification of anthropogenic relief changes, which is a considerable challenge in regions with historic mining activities.

tification of anthropogenic relief changes, which is a considerable challenge in regions with historic mining activities.
Here, the Ville (Rhenish lignite district, Germany) is used to quantify surface mining induced relief changes in one of the oldest and currently largest lignite districts in Europe. Historical maps from first geodetic mapping in 1893 enabled construction of a historic digital elevation model to quantify the relief changes in comparison to elevation data from 2000 and 2015. The vertical accuracy of the historic data is remarkably high, with relief differences < 2 m in areas not affected by mining. In total, 49.2% of the investigated area (184 km 2 ) shows a relief deficit and 14.5% has positive relief differences. Absolute changes account for more than 80 m heightening (dumpsites of overburden) and lowering of the natural relief (pits). Besides these altitudinal changes, overall steeper slopes are significant for the new topography, but levelling exists likewise. The spatial variabilities are discussed in the context of the regional geology and the mining techniques.
Undoubtedly, such large-scale anthropogenic relief changes persist for a very long time and will last as a human legacy far into the future. Only the detailed reconstruction of the pre-mining relief offers the ability to clarify the dimension of humans as geomorphological agents and to understand landscape perception. Due to the fact that the impact of open pit mining has such a large vertical and horizontal extension, their consideration as part of anthropogeomorphology can significantly contribute to support future Critical Zone research in the Anthropocene.

K E Y W O R D S
Anthropocene, anthropogeomorphology, digital elevation model, historic maps, mining 1 | INTRODUCTION Geomorphological investigations can detect, quantify and monitor not only natural but also anthropogenic relief and landscape changes and can therefore contribute to the actual discussion about the "Anthropocene" (cf. Brown et al., 2017). In this context, the importance of geomorphology can be summed up in the statement of Brown et al. as 'Rather than being drawn into stratigraphical debates, the primary concern of geomorphology should be with the investigation of processes and landform development, so providing the underpinning science for the study of this time of critical geological transition' (Brown et al., 2017, p. 71). Today it is accepted, that humans, as geological and geomorphological agents are nowadays responsible for a higher amount of sediment and rock movement than the sum of all-natural processes (cf. Hooke, 2000;Price et al., 2011;Aguilar et al., 2020). Therefore, anthropogeomorphology the 'study of the human role in creating landforms and modifying the operation of geomorphological processes' (Goudie & Viles, 2016, p. 7) has subsequently emerged as a major field of research regarding any concerns about geomorphology within the Anthropocene. One of the very obvious and direct/intended anthropogenic landscapes, where humans have shaped the natural surface of the earth tremendously, are regions with surface mining activities. Very recently, the first disclosed global database detected more than 21,000 areas of active surface mining equivalent to an affected area of around 57,000 km 2 (Maus et al., 2020). Surface mining creates new negative landforms caused by excavation, new positive landforms due to dumpsites, and levelling landforms (cf. David, 2010;Mossa & James, 2013;Tarolli & Sofia, 2016). Other studies emphasize the importance of autonomous extraction of anthropogenic topographic signatures in mining landscapes in order to mitigate geomorphic hazards and sustain environmental planning (e.g., Chen et al., 2015). It represents a trend to use high-resolution digital elevation models (DEMs) to derive human topographic signatures in landscapes, where also ongoing acquisition of recurring high-resolution topographic data are used for active monitoring of surface changes and processes (Tarolli & Sofia, 2016;Tarolli et al., 2019). Given this modern ongoing data availability, historic quantification of anthropogenic changes over decades and centuries is limited due to missing historic DEMs.
The implementation of historical maps however, facilitates the inclusion of an extended temporal dimension, if an accurate historical DEM can be computed based on them. This serves as a baseline for the computation of DEM of differences (DoD) (James et al., 2012). This is of particular importance in the investigation of anthropogenic relief changes in old mining regions. Here, the strong human impact was established in historic times. Some of the previous mining activities have been completed and subsequent land reclamation has been accomplished.
Recent research from the German working group on geomorphology demonstrates how historic DEMs can reconstruct the pre-Roman topography for the city of Aachen (Pröschel & Lehmkuhl, 2019), the palaeorelief of Leipzig from the year 1050 (Grimm & Heinrich, 2019) or the pre-modern terrain for the Charlemagne's canal construction site in Bavaria (Schmidt et al., 2018). In these examples, data from archaeological excavations and geological drillings have been used. Harnischmacher and Zepp (2014) display the impact of anthropogenic induced subsidence due to underground mining along the Ruhr District based on elevation data from first geodetic mapping represented on historical maps from the Prussians (Preußische Neuaufnahme).
However, the analysis of area-wide anthropogenic relief changes on the basis of historical topographic maps of large surface mining areas, which differ obviously from that of underground mining areas, does not presently exist.
Following this context, the study at hand aims to detect and quantify human induced relief changes in the Ville mountain range, a tectonic horst structure in the Cologne-Aachen area, Germany (southern Lower Rhine Embayment). Here, substantial anthropogenic relief changes, not only with regard to extensive open-pit lignite mining but also with regard to reclamation and recultivation, are evident over the past 200 years. Since the Lower Rhine Embayment comprises fertile loess soils, the former mining area received very early attention from international specialists for the process of reclamation (e.g., Elkins, 1953;Nephew, 1972).
Issues with regard to resulting landscape changes focusing on relief differences are: (i) in which dimension changed the relief under the influence of mining, (ii) which new relief forms were created, and (iii) what is the relationship between new landforms and land use?
Altogether, this can be achieved by a detailed geomorphological mapping created by an adequate data background in spatial and temporal scales for the period of interest.

| STUDY AREA
The Rhenish mining area extends over an area of 2500 km 2 (Dworschak & Rose, 2014). It represents the largest lignite district in Europe and the main focus of lignite mining in Germany (95 Mt production/a), in addition to the Lusatian mining area in eastern Germany (62 Mt production/a) and the central German mining area (19.3 Mt production/a) (Alves Dias et al., 2018). Together with Poland, Bulgaria and Romania, Germany has also the largest lignite mines in the European Union, predominantly operating as open-pits. In total, the three lignite regions in Germany contain more than 10% of the remaining global commercially exploitable brown coal deposits (Meschede & Warr, 2019) and Germany is currently the second largest coal producer in the world, following China.
In this context, the Rhenish mining area within the southern Lower Rhine Embayment in the Cologne-Aachen region (

| Tectonic setting and actual land use
The southern Lower Rhine Embayment is part of the European Cenozoic rift system, where tectonic subsidence occurred, since the Oligocene and lignite deposits originate from Miocene peat deposits. Two major seams (main-and upper-seam) have thicknesses between 10 m and up to 40 m respectively and exist next to various thinner peat deposits [cf. Hager (1993) for the origin of lignite deposits in the lower Rhine region]. The ongoing tectonic activity of the Lower Rhine Bay has divided the basin into different separate blocks along the main trend lines from northwest to southeast. The depth of the lignite deposits below the present surface is dependent on the different tectonic subsidence and erosion/deposition of Pliocene and Quaternary, mainly unconsolidated sediments (cf. Ahorner, 1962). Apart from minor Late Pleistocene loess deposits, the Middle Pleistocene upper terrace of the Rhine River represents the natural surface of the Ville horst with varying thicknesses (Boenigk & Frechen, 2006). In the southern part of the Ville horst, the lignite occurs near the surface and subducts towards the north, partly to more than 450 m below the surface (up to about 370 m below sea level).
Palaeoseismology and modern-day seismicity verify active tectonics in the Lower Rhine Embayment in the Holocene, including historic times (compare latest summary in Hürtgen et al., 2020). For example, a strong earthquake (ML 6.1) occured in Düren in 1756 (Ahorner, 1962), located approximately 20 km west of the Ville horst. Active faults are also situated just 5 km south of study area (Kübler et al., 2017). In addition, recent movements detected by geodetic measurements point to mining-induced subsidence of up to 2 cm/yr due to groundwater extraction, but also due to ongoing extension of the rift system with a few millimetres per year within the Lower Rhine Embayment (Campbell et al., 2016). Thus, the complex tectonic setting is not only responsible for the spatial distribution of the lignite seams, but also is an important impact factor, which could affect recent and future anthropogenic landforms.
As the study area was chosen to demonstrate changes between the natural landscape and recultivated post-mining landscape, its current land use does not indicate the landscape history at first glance.

| Mining history and excavation techniques at a glance
The presence of lignite may have been known since Roman times, when Tacitus describes burning soils in AD 59 (Kaever, 2004). First documented industrial mining activities occurred in the late 19th century with the introduction of industrial lignite briquettes (Brunotte et al., 1994). Mining began in the southern part of the Ville, where the Tertiary lignite layers were exposed near the surface. This resulted in the excavation of very small and shallow mines and only small amounts (< 10 m) of overburden material (sands and gravels) had to be removed. Here, no fertile soils were destroyed and only woodland was affected by mining (Dickmann, 2011). Thus, there were almost no land-use conflicts at this 'island of infertility' (Elkins, 1953, p. 132).
The first General Mining Act existed in 1865 and had only very few statements on recultivation and reclamation of mining areas (Dickmann, 2011). With regard to the tectonically submerging lignite layers towards the north, technical developments in the first half of the 20th century allowed deeper excavation and the further expansion of mines. For the purposes of land reclamation, forestry and farming recultivation, the growing amount of overburden from unconsolidated sediment masses, were to be deposited in mined out pits (state mining authorities in 1929, cited from Dickmann, 2011). Since the 1950s, the introduction of the first bucket-wheel-excavator opened up pits with operating depths > 200 m below the natural surface. These were the largest open pit mines in the world at that time (Bauer, 1971). Increased mine depth resulted in increased overburden material which in turn led to too little deposit capacity in the shallow mines. The material was consequently dumped at heap sites. One example was Glessen. It was built between 1955 and 1970 and has a height of 80 m above the natural landscape and has a storage capacity volume of 170 million m 3 (Brunotte et al., 1994). Hence, the thick layers of overburden, unconsolidated sediments, which prevent dense contour lines. Two parameters need to be set before computation: appropriate pixel size of the DEM and the interpolation method.
Different approaches for the use of an appropriate pixel size are given in Table 1.
As the input data itself have a fixed spatial resolution given on the map scale (here 1:25,000), a first approximation for an appropriate pixel size can be calculated from a cartographic point of view. Tobler (1987) suggests a maximum spatial resolution with the division of the denominator of the map by 1000 and a minimum spatial resolution half of this amount, which lead to a pixel size between 12.5 m to 25 m for our input data. Hengl (2006) sees the coarsest grid resolution based on cartographic concepts in relation to the minimum legible delineation as scale number (SN) * 0.0025, which results in a maximal spatial grid resolution for the Preußische Neuaufnahme of 62.5 m.
The finest spatial resolution is depending on the maximum location accuracy and results in 2.5 m pixel resolution for a map scale of 1:25,000 (SN * 0.0001). The minimum spatial resolution is calculated with pixel size = total size of study area/(2 × total cumulative length of digitized contours) (Hengl, 2006). In this case, the final DEM needs a minimum pixel size of 45 m [p = 184 km 2 / (2 * 2030 km)]. This is in accordance with the average distance from any point in the study area to its next contour line, which is 43.3 m. The most distant unknown point is 426 m away from its next contour line. Hence, a minimum pixel resolution of 43.3 m is needed in average to display height differences between two contours accurately. The mean distance from an unknown point to its next contour line is within the 20% of most dense region of contours, thus the area of most relief changes per area is 13.1 m and represents the maximal pixel resolution with regard to contour density. A stricter approach based on the derivation of the 5% probability smallest width of contours (Hengl, 2006) is avoided as the internal error of the input data itself can have an impact on this very strict approach. Thus, the region with 20% of the densest contours is a more reliable threshold for this study. Ultimately, a pixel size of 30 m was chosen for the calculation of the 1893_DEM, which fulfils all above criteria.
The second important parameter for the computation of the 1893_DEM is the interpolation method. Here, a Thin Plate Spline function (Donato & Belongie, 2002) has been used for the interpolation of the historic DEM using SAGA GIS Software (Conrad et al., 2015). Input parameters were set to a maximum search distance of 450 m (maximal distance to an unknown point), local search range and a minimum of 16 points for interpolation. This accounts for the prevention of over-shooting and over-smoothing as a possible problem with spline interpolation (Hengl & Evans, 2009).  The largest calculated catchment counts to 15.7 km 2 ( Figure 2C) and

| Slope differences 1893-2000-2015
The slope classification for the natural topography of the Ville in 1893 has its maximum with 43.7% of the area ranging between > 0.5 and 2 (Table 2). Only 2.4% shows values in the classes of 7 to 15 .
Slopes with values above 15 are absent.
The spatial distribution of slope values ( Figure 4A

| Quality and comparability of the datasets
The accurate computation of the historic DEM is crucial for the quantification of the anthropogenic topographic impacts in the area under investigation. The used spline algorithm avoids a linear interpolation (Reuter et al., 2009). Systematic errors are often represented by terraces, a common feature in interpolated DEMs based on contour data (Reuter et al., 2009). The high correlation, over the total study area, between the input heights (digitized from the contours) and computed heights indicated with the very small root mean square error (< 1 m between observed and predicted height) shows that the spline algorithm did not produce large over-shooting and interpolation errors with the given input data parameters. However, no information exists regarding the vertical accuracy of the mapped contour lines, as the height information itself is only represented as an absolute value on the analogue maps without metadata. This is a common problem for the use of contours from analogue maps and other studies, for example Harnischmacher and Zepp (2010) account for this with a nugget variation of zero in their interpolation of the historic DEM using the kriging interpolation. This accounts for the missing vertical error assessment of the input data. A reference area in the north-eastern part of the Ville horst, where no mining or other larger human activities took place, constitutes for the accuracy assessment of the historic T A B L E 2 Slope classes according to low land classification of Leser and Stäblein (1975)  F I G U R E 4 Slope classification with classes according to low land topography in geomorphological mapping (Leser & Stäblein, 1975) for (a) the historic DEM (1893) Overall, the data accuracy of the historic DEM from this study is comparable to the very few other studies, which have used the Prussian topographic maps previously with regard to geodetic information. Our threshold is comparable to the accuracy of the historic DEM in the Ruhr District, where 51% of data are affected with an error between 1 and 2 m, 48% below 1 m and only 1.3% with an error above 2 m (Harnischmacher & Zepp, 2014). Similarly, the historic DEM based on the Prussian topographic maps (here from 1883) for the Upper Silesian Coal mining area has a maximum altitude error of ±0.45 m (Dulias, 2016). Whereas these studies used the Prussian topographic maps mainly for the reconstruction of a historic DEM to map relief subsidence due to underground coal mining, another example based on these maps is the reconstruction of coastline changes, where the maps have a horizontal average accuracy of 5 m (Deng et al., 2017).
Apart from the approximation to get the right pixel size with the coarsest and finest legible raster resolution for the given input data, 'No absolute ideal pixel size exists, that is for sure' (Hengl, 2006(Hengl, , p. 1297. In general, varying density of points and a generalization of topography challenge the interpolation of a gridded DEM derived from digitized contours (Reuter et al., 2009) Eurasia is on average ±6.2 m for the absolute height (Rodriguez et al., 2006), which is comparable to the accuracy assessment of the SRTM-data in comparison to the German DEM50 in a study area southeast of Cologne (Bolten & Bubenzer, 2006). In comparison, the LiDAR data have a vertical accuracy of ±20 cm (Geobasis NRW, 2019). Whereas the LiDAR data from 2015 are able to dissolve tree canopy heights, the SRTM-data are more affected by errors in for-

| New anthropogenic landforms
Apart from the absolute relief differences, the character of the postmining topography has significantly changed the landscape setting.
This is remarkably visible in the direct comparison of the topography between 1893 and 2015 ( Figure 5). The anthropogenic relief is much more pronounced and more heterogeneous than the original natural relief and constitutes a typical human topographic signature for relief alternation (Tarolli & Sofia, 2016).
A central question is if the anthropogenic topography has a distinct morphometric fingerprint (cf. Tarolli  wide distribution of slope ranges (e.g., Leser & Stäblein, 1975). The increase of higher slope classes as an effect of the DEM with a higher resolution can be excluded, because the slope classes for 2015 are also calculated based on the aggregated DEM with 30 m pixel resolution. The initial spatial resolution of 1 m shows 10.8% of the total area with slopes higher than 11 . It is remarkable that the general decrease of slope from its maximum between > 0.5 and 2 up to >11 -15 is interrupted by an increase of 7.8% with slopes > 15 (cf. Table 2). Here, the lowering of topography has a similar anthropogenic signature as mountaintop mining (Ross et al., 2016). In addition, reclamation for new agricultural landscapes has also levelled the topography and produced a more gentle relief with only very weak inclination for surface runoff (Dworschak & Rose, 2014).
Overall, the spatial differences and distribution of total relief changes return to the geological structure of the Ville and the technological mining history from small-scale surface mining in the south without any larger pit sites, to large-scale deep mines in the central and northern part. The latter will be subsequently discussed in more detail.

| The southern part of the Ville
The first open pits for lignite mining were mapped in 1893 (Preußische Neuaufnahme) in the southern part of the study area.
However, the limited extent of these pits in their initial status did not significantly change the topography and there are no artificial dumpsites in the surrounding of these pits, due to the thin sedimentary cover above the former lignite layers. Such mining activities took place in times with no detailed regulations for post-mining reclamation and recultivation. For example, narrow linear 'valleys' represents former railroads for the lignite transport, which were excavated subsequently. The resulting small-structured relief was unplanned and the small pits were scarcely recultivated or refilled. This resulted in the formation of various lakes ('Ville lakes'). However, in comparison to the pre-mining landscape, the heterogeneous and only partly experimental recultivation of the southern part during the first reclamation phase  resulted in a higher ecological diversity with new biotopes, plant and animal communities (Bauer, 1971). Given the fact, that this region was formerly seen as poor-quality land with regards to agriculture (Elkins, 1953), the enhanced ecological diversity and the post-mining land use has provided local recreation areas for the cities of Cologne and Bonn with forests and small bathing lakes. They are examples for the positive potential of post-mining landscapes. (Wirth et al., 2017). To our knowledge, this region constitutes the earliest examples, where already during the 1950s the conscious transformation of a post-mining landscape as a recreation area for the local inhabitants was described (Elkins, 1953) and executed.
The generally lowering of the relief, without any larger anthropogenic landforms and only small-scaled more heterogeneous landforms, becomes clear in the detailed comparison between 1893 and 2015 ( Figure 6).

| The central and northern part of the Ville horst
The central and northern part of the Ville host is characterized by mines of greater depth and dimension. Here, in contrast to the sparsely populated infertile southern part, farmland and villages with several thousands of inhabitations had to be relocated (Dickmann, 2011). This resulted in a stronger focus on planning legislation, which also affected the created anthropogenic relief. The location of the Glessener Height is the largest dumpsite in the investigated area was planned to be built as high as possible so as to minimize the amount of land required. This resulted in the very steep, forested slopes as well as the flat top designated as agricultural fields ( Figure 7A). As there were no existing large-scale open pits before, in which to deposit the thick overburden sediments from the first deep mines in the northern part of the study area, this type of landscape and relief are typical for the reclamation period of the 1950s and 1960s (cf. Brunotte et al., 1994).

| The Ville as an Anthropocene landscape par excellence
The basic consideration of anthropogenic landforms as study objects in anthropogeomorphology is obvious, but can also be integrated into the Anthropocene idea and the 'Critical Zone' conceptual approach (Aguilar et al., 2020: 1). In this context, anthropogeomorphological studies are able to meet the need and challenge for broad interdisciplinary and transdisciplinary studies. The transformation of mining areas, although complex, can also be researched with this approach, in which the entirety from the vegetation cover to deep groundwater changes has to be regarded (Giardino & Houser, 2015 Brunotte et al. (1994) argue, that all high dumpsites will stay as foreign objects in the landscape of the Lower Rhine Embayment. It is in general a common feature that mined landscapes are readily recognizable (Mossa & James, 2013).
However, undoubtful from a geomorphological knowledge and perspective, it will be interesting how future generations, especially with no related expertise, will take note of the anthropogenic landforms.
Currently, people with the experience of active mining are still living in the area. However, more and more inhabitants without such experience are present. Often, the current land use with forest (33.6%) and agriculture ( for the major impact of mining apart from the morphological changes is almost absent in some regions. Hence, the reconstruction of the natural relief of the Ville is also an important resource to visualize the actual natural (here seen as pre-mining) landscape to be aware of the tremendous changes through human activity.

| CONCLUSION
In order to facilitate a detailed reconstruction of the pre-mining relief, historic maps (if available) are one of the most valuable sources, especially for comparisons with actual satellite or geodetic data. Here, the historic map of the Preußische Neuaufnahme from 1893 represents the natural topography of the oldest part of the current largest lignite district in middle Europe, the Ville horst in the southern Rhine Embayment. The focus of this contribution with a comparison of a computed historic DEM with the modern topography delimits our approach from other studies with their focus on the derivation of mining induced topographic signatures based on morphometric parameters of the nowadays relief (e.g., Chen et al., 2015;Tarolli et al., 2019;Cao et al., 2020).
The Ville area represents one of the oldest and largest anthropogenic mining landscapes worldwide. As reclamation is still a nascent discipline (Mossa & James, 2013), the region provides an excellent example to gain a better general understanding of long-term anthropogenic processes and their environmental impacts. The created historic DEM set the baseline for the mining induced morphological changes and can further be extended to the ongoing mining regions, but also to minimize ongoing disadvantageous earth surface processes, which is an ongoing challenge in geomorphology, respectively anthropogeomorphology (Tarolli & Sofia, 2016).
Given the size of the study area and the scale of interest, the produced DEM (30 m spatial resolution) is sufficient enough to represent the natural topography of the Ville before the mining and the associated relief differences in comparison to the relief after recultivation.
Hence, the approximation for a suitable pixel resolution based on the contour spacing depending on their cumulative length over the study area and their maximal density as indicator for the complexity of terrain under investigation is a valuable approach. The results show that the vertical dimension in which the relief has transformed is at the scale of several decametres, both in positive and negative direction.
New created landforms range from small-scaled heterogeneous and unplanned landforms in the southern part of the Ville, whereas large positive dumpsites with steep slopes characterize the northern part of the study area. The differences between the regions originated in the natural geological setting of the lignite strata but also in the technohistorical development in mining and recultivation through time.
The regional scale of this study can be used for subsequent research questions at the local scale, for example for morphotectonic or hydrological changes of single catchments. It could be possible to investigate the natural topography at a higher spatial resolution in selected areas, as the initial point density varies over the study area and subsections could be expanded. Most scientific studies within post-mining landscapes focus on consideration of landscape functionality, ecology (including artificial soils) and subsequent land use, rather than the relief itself. However, primary geomorphological considerations should serve as a foundation, as many of the subsequent topics are significantly influenced by the relief. Thereby, anthropogenic geomorphology is of particular importance to clarify the human impact on landscapes, also for non-experts. Formerly mined and today recultivated or reforested areas do not always establish a direct connection for an indication of their enormous historic anthropogenic transformation. However, such changes will undoubtedly persist for a very long time and will therefore last as a human legacy far into the future, so that anthropogeomorphology can significantly support future Critical Zone research in the Anthropocene. In addition, a recommended extension of the results with regard to socio-cultural purposes can substantially contribute to the discussion about the Anthropocene in social sciences. Finally, an integration of questions about the perception of landscapes can significantly enhance a geographically human-environment perspective.