476 Documenta Praehistorica XLVII (2020) Introduction Remote sensing techniques and Geographic Informa- tion Systems (GIS) have proven to be useful tools in environmental research, and particularly in model- ling supraregional surface developments and land- cover changes (Kaplan, Avdan 2017; Landuyt et al. 2019; Malekmohammadi, Jahanishakib 2017; Shen et al. 2019). Open source medium-resolution satel- lite images from the Landsat and Sentinel missions are used to monitor and map surface cover modifi- cations through multispectral analysis (Stratoulias et al. 2018). These methods are extended by active sensor radar analysis that allow for surface observa- Fables of the past> landscape (re-)constructions and the bias in the data ABSTRACT – Prehistoric landscape reconstructions are still considered an unsolved methodological issue in archaeological research, and this includes the perception and transformation of an indivi- dual landscape in relation to situational and local ecosystem performances. Which parts of the land- scape offered the potential for land-use and which areas were rather unsuitable due to a variety of environmental preconditions? The modern perception of the archaeological record that is distri- buted in the modern landscape does not necessarily represent a realistic dispersal of past human activity, but rather reflects the current state of archaeological research and modern land-use strate- gies. This contribution provides a critical assessment of spatial analyses of large and unstructured archaeological datasets and the non-reconstructibility of past, individually perceived palaeolandscapes. IZVLE∞EK – Rekonstrukcije prazgodovinske krajine ∏e vedno veljajo za nere∏eno metodolo∏ko vpra- ∏anje v arheolo∏kih raziskavah, kar vklju≠uje zaznavanje in preoblikovanje posamezne krajine glede na situacijske in lokalne u≠inke ekosistemov. Kateri deli pokrajine nudijo potencial za izrabo zem- lji∏≠ in kateri predeli so zaradi razli≠nih okoljskih danosti manj primerni? Sodobno dojemanje ar- heolo∏kih zapisov, ki so raz∏irjeni v sodobni krajini, ne predstavlja nujno realne razpr∏enosti ≠love∏- kih aktivnosti v preteklosti, temve≠ odra∫a trenutno stanje arheolo∏kih raziskav in sodobne strategi- je rabe krajine. V prispevku nudimo kriti≠en razmislek o prostorskih analizah velikih in nestruktu- riranih arheolo∏kih podatkovnih baz in neobnovljivosti preteklih, posami≠no zaznanih paleokrajin. KEY WORDS – spatial analyses; GIS; multivariate modelling; landscape archaeology; human ecology KLJU∞NE BESEDE – prostorske analize; GIS; multivariatno modeliranje; prostorska arheologija; ≠love∏ka ekologija Bajke o preteklosti> krajinske (re)konstrukcije in pristranskost podatkov Michael Kempf Department of Archaeology and Museology, Masaryk University, Brno, CZ Physical Geography, Institute of Environmental Social Science and Geography, Faculty of Environment and Natural Resources, University of Freiburg, Freiburg, DE Archaeological Institute, Faculty of Humanities, Dep. Early Medieval and Medieval Archaeology, University of Freiburg, Freiburg, DE kempf@phil.muni.cz< Michael.kempf@archaeologie.uni-freiburg.de DOI> 10.4312\dp.47.27 Fables of the past> landscape (re-)constructions and the bias in the data 477 ries, resulting in an increasingly blurred terminology that makes it difficult to understand the methodolo- gy and its limitations (David, Thomas 2010; Meier 2017). Landscape archaeology has a rather short hi- story, and only came into use in the mid-1970s (Da- vid, Thomas 2010; Fleming 2006). Moreover, it took until the 1980s and Colin Renfrew’s advance in the field of cognitive archaeology for landscape archae- ology to become established in post-processual ap- proaches (Doneus 2013): the categorical separation or inclusion of culture and environment (Ingold 2000; Meier 2009). Basically, landscape archaeolo- gy has now become an umbrella term for spatial pat- terns in archaeology (Doneus 2013). It aims to un- derstand how space has been organized and struc- tured in premodern societies through the emotional meaning, experience and categorization of land- scapes (Meier 2009). Landscape archaeology thus does not simply represent an extension of environ- mental or settlement archaeology, but in contrast, a conglomerate with explicitly cultural-scientific me- thods for the social reconstruction of spatial life worlds (Meier 2017). The integration of trans-regio- nal geographic networks and the dissolution of the local environment enable an objective consideration of resource distribution, land-use, supraregional com- munication, mobility and exchange, as well as tran- scultural adaptation and development processes. Human ecology and the spatial-temporal scale of landscapes affordances Beside a conceptual framework of archaeological cri- teria to define spatial patterns of human behaviour, landscapes are considered as being composed of many characteristics. Michel Baguette et al. (2013) consider the landscape as the most appropriate spa- tial scale to define ecological networks in ecosystems. According to the authors, the extreme difference in the perception of the term landscape emerges from the divergence of the concepts of biogeography and behavioural ecology. Biogeography defines the land- scape as a clearly categorized spatial organization with a homogeneous geomorphology and climate. Behavioural ecology, on the other hand, defines the landscape as the individual’s perception of the en- vironment and the spatial extent of his/her activi- ty range as a function of the lifetime spread of the organism (Baguette et al. 2013; Gurrutxaga et al. 2010; Kupfer 2012; Schaich et al. 2010). It is obvi- ous that landscapes can hardly be defined solely through spatial determination of human-environ- ment interactions without adding a temporal com- ponent and an individual dimension of landscape perception. tions during cloud-cover or without sunlight (SAR – Synthetic Aperture Radar) (Cao et al. 2019; Dabrow- ska-Zielinska et al. 2016; Landuyt et al. 2019; Mlecz- ko, Mróz 2018). The massive anthropogenic pres- sure on today’s ecosystems drastically expands the need for large-scale surface monitoring. This is par- ticularly visible in the growing social and cultural vulnerability to extreme weather events, which re- quires the intensification of large-scale surface moni- toring to understand the relationship between natu- ral ecosystem impacts and cultural heritage man- agement. The integration of landscape connectivity, the human as vulnerable agent, and increasing eco- system susceptibility plays a key role in landscape archaeological research (Kempf 2019b; Lasapona- ra, Masini 2006; 2011; 2013; Masini, Soldovieri 2017; Morrison 2013). The evaluation of the distri- bution of archaeological sites and human behaviour in a specific landscape demands deeper knowledge of the geographical interconnectivity of the environ- mental preconditions. Intense land-use and settle- ment activity in particular severely modified the earth’s surface in past centuries, building a variety of cultural landscapes on top of each other. It is a methodological challenge to evaluate the patterns and structures behind the distribution of archaeolo- gical sites in the landscape, in order to rapidly draw conclusions about landscape permeability, cultural exploitation, and the human-environment interac- tion of premodern societies. This contribution aims to highlight the interface between the monitoring of surface dynamics, the reconstruction potential of palaeoenvironments, and the analysis of spatial pat- terns of archaeological site distribution. The follow- ing questions are of central importance in this con- text: ❶ How is our modern understanding and percep- tion of an archaeological landscape biased by mo- dern land-use concepts, settlement activities, and recent structural surface changes? ❷ How can GIS-based environmental models, remote sensing applications, and statistical analysis ex- plain spatial patterns of archaeological site distri- bution? ❸ How can these concepts contribute to a compre- hensive landcover reconstruction? Fables of the reconstruction? Landscapes, eco- systems and affordances Landscape archaeology is in vogue, and there are in- creasing discussions about the terminology of land- scape. This has led to a mixture of concepts and de- finitions from many scientific fields and subcatego- Michael Kempf 478 The temporal component is a methodological con- fusion in landscape archaeology. This is particularly important in terms of the differentiation of event and process. Events seem to take place on a short- term scale with noticeable and mostly severe impacts on ecological habitats and sociocultural human sys- tems (Berglund 2003; Büntgen et al. 2011; Toohey et al. 2016). However, the differentiation between event and process in archaeology is more deter- mined by the material consequences than the envi- ronmental triggers. As a result, short-term events tend to blur in long-term chronological categoriza- tion. They are not detectable until their consequen- ces are not manifested materially, socially and cultu- rally. Events, processes, and the spatial parameters of landscape patterns and susceptibilities are inevi- tably linked, and form the specific dynamic charac- ter of landscape ecology and archaeology. Tracing where and when groups and individuals have settled and reshaped a particular place for a certain reason is of central importance in archaeolo- gical research, and especially in cultural heritage management (van Leusen, Kamermans 2011; Ver- hagen et al. 2010; Verhagen 2018). The basis for this is relatively simple: human behaviour is patterned (Brandt et al. 1992). The resulting structures follow the conceptual landscape fragmentation of premod- ern societies, and eventually their interaction with their environment (Verhagen 2007). This geograph- ic fragmentation and the patterned human behav- iour are strongly connected to the concept of so- called landscape affordances. The neologism affor- dance, first introduced by James Gibson in the late 1970s, describes the phenomenon of propositions emanating from objects within a specific environ- ment (Gibson 1979; Jung 2018; Loveland 1991). Affordances are not (meta)physical properties, but rather empirical meanings that are in some way ar- ranged in space (Jung 2018). Affordances were first introduced into archaeological discourse by Timothy Ingold in 1992 (Gillings 2009; Ingold 1992; 2000). In contrast to defining the components of the envi- ronment as passive resources, the concept of land- scape affordances aligns dynamic and processual feedback with an individual’s behaviour in the mo- ment of mutual interaction (Gillings 2009). In a broader sense, these fundamentals are decisive for the differentiation of landscape and environment, which Ingold characterizes through objective and subjective or internal and external observers (Ingold 2000; Meier 2017; Webster 1999). Affordances are not a universal concept for certain actions of social groups with material objects or elements in their environment, but take place at the individual level of perception of an object in the immediate moment of its confrontation. According to David Webster (1999), the relationship between affordances and landscapes can be divided in two: low-order invari- ants denote the individual elements of a landscape, while high-order invariants summarize these ele- ments and generate potentially available/not avail- able or usable/not usable surfaces that are offered to an individual. Furthermore, Mark Gillings (2007) describes affordances as disposition properties which can be divided into direct and potential compo- nents. Nevertheless, both characteristics constantly coexist. Although the concept of affordances is much older than the basic idea of GIS-based multivariate land- scape reconstructions in archaeological research, both systems consist of similar components: the se- lection and categorization of environmental parame- ters and preferential sites in relation to the personal interests and actions of individuals in their environ- ment. Preferences in land-use are not only physical interrelations between the needs and demands of people and their surroundings. According to Marcos Llobera (1996; 2001), changes in affordances reflect social changes within a group. Individuals in a parti- cular group share common or similar structures, de- velop similar practices, and consequently share simi- lar affordances. A possible method for the reconstruction of human patterns in the landscape is the application of mul- tivariate modelling. In landscape archaeology, multi- variate modelling is based on the integration of a va- riety of GIS-based datasets (Groenhuijzen 2019; Ho- wey 2011; Howey, Brouwer Burg 2017; van Dinter 2013). The inductive approach of multivariate land- scape models is the recognition of specific location parameters in the archaeological dataset (Güimil-Fa- riña, Parcero-Oubiña 2015; Weaverdyck 2019). Di- gitally obtained integrative accumulative surfaces allow for the evaluation of environmental parame- ters without completely excluding human interac- tions. The diachronic reflection of the archaeological record of a study area helps to identify patterns and continuous human impacts on the landscape on large temporal and spatial scales. Anthropogenic surface modifications – how modern is the past? The French part of the Upper Rhine Valley was cho- sen as the study site. The area covers about 8300km2 with large-scale geographical feedback and ecosys- Fables of the past> landscape (re-)constructions and the bias in the data 479 tem connectivity (Kempf 2019b). In order to eval- uate the natural conditions of the study area, the actual environmental conditions and the recent sur- face changes were modelled on the basis of histori- cal maps, modern satellite images and various GIS- attributes and datasets. The whole region was mas- sively modified by intensive land-use and increas- ing construction development within the past few decades. Climatic extreme events and long-term va- riability have also triggered droughts, flooding, and surface transformation (Giacona et al. 2018; Glaser et al. 2010; 2012; Himmelsbach et al. 2015a; 2015b). The lowlands in particular are prone to increased temperatures, heat waves, and drought stress (Duch- ne, Schneider 2005; Muthers et al. 2017). For prehistoric societies it was periodic events that especially shaped perceptions and opportunities in the landscape. This means that the sum of spatial re- quirements is determined by the vulnerability of the environment to extreme events and the maximum benefit that can be assumed with an acceptable risk of loss. This results in a long-term trend in land-use which does not define areas of high suitability ac- cording to qualitative and modern standards, but is formed by periodic empirical values. Do pre-modern landscapes largely consist of experiences that are no longer accessible today? If this is the case, then the question arises to what extent today’s surfaces are still parts of the physically existing landscapes of pre- modern societies, and how much palimpsest is still present in the landscape? A review of the environ- mental variability over the last 150 years is enough to identify the massive interventions in the ecosys- tem’s balances. Large-scale infrastructure develop- ment, new urban areas, deforestation and expansion of arable land, drainage and exploitation of resour- ces are just a selection of the anthropogenic impacts on the land surface. The rapid change in landcover can be tracked by comparing historical maps, mod- ern satellite images from different years, and more recent landcover data sets such as Corine Landcover (CLC). Material and methods A Geographical Information System (GIS) is more than just a simple software tool for storing and ma- nipulating spatial data. Much of the actual work that happens before visualization, spatial analysis and database management is the acquisition of spatial data that fits the desired spatio-temporal resolution of the research framework. The issues that were raised by the increasing application of GIS in inter- disciplinary research led to the distinction between GIS (software tools) and GISc (Geographic Informa- tion Science), with the latter concerned with the many conceptual interrelationships between science and the humanities (Conolly, Lake 2006). Multiva- riate landscape analyses are based on selected and hypothetical environmental parameters. The selec- tion of the parameters is carried out empirically via the feedback mechanisms of an ecosystem. For exam- ple, the type and composition of quaternary sedi- ment stratigraphy in connection with groundwater height, flood risk and average precipitation rates de- termine soil formation processes and small-scale soil mosaics. From the estimation of the numerous (mul- tivariate) determinants, a potential premodern land- scape can be deduced (Fig. 1). In reality, however, this surface is based on modern empirical data and can only be transferred to prehistoric surface forma- tions with considerable uncertainties. Nevertheless, these models allow us to draw conclusions about po- tential prehistoric land-use concepts, because they integrate ecosystem connectivity on both the small and the large scales. A broad variety of spatial and temporal environmental datasets have been acquir- ed, manually developed or processed from various departments, institutions or through open source online portals. One major difficulty for the current study was the synchronization of datasets from the French and German sides of the Upper Rhine Valley, which have different geographical coordinate sys- tems, spatio-temporal data resolution, typology, and particularly data availability and accessibility. The following descriptions list the respective datasets and briefly summarize the methods and strategies of the digital manipulations for the study area. Environmental conditions A comparison of a historical map from the 19th cen- tury with images from two satellite missions (Land- sat-1, sensing date 9th October 1972; Landsat-OLI8 sensing date 24th September 2018) shows signifi- cant transformations of the surface cover during the past 150 years (Fig. 2). Massive deforestation activ- ity took place that aimed to transform the surface into arable land or to be suitable for use as con- struction sites for increased urban and rural devel- opment. However, in order to understand the distri- bution of the archaeological sites in the landscape and consider the potential movement behaviour of past societies, landscapes need to be differentiated into their physical parameters such as climate, geo- logy, and hydrology, and into their artificial and cultural components based on anthropogenic im- prints. Michael Kempf 480 Two surface classifications can be deduced from the evaluation of landcover changes and land-use: a land- scape suitability model and a landscape bias model that evaluates the impact of modern surface trans- formations. These surfaces include geological units, soil quality and drainage potential, flooding vulne- rability, groundwater level, and historical surface dynamics, such as infrastructure change, settlement expansion and modifications of the hydrological sys- tem. Based on these potential maps, archaeological and modern land-use patterns can be quantitatively compared and tested for their spatial interrelations. The environmental factors have been analysed on the supraregional scale to identify the large-scale con- nectivity patterns of the Upper Rhine ecosystem. The geological and pedological data that supports the analyses of the study site were acquired from the Bundesanstalt für Geowissenschaften und Rohstoffe Hannover (BGR). For the French part of the Upper Rhine Valley, soil maps from the ARAA (Association pour la Relance Agronomique en Alsace, http://www. araa-agronomie.org/, last accessed 19th January 2019) and the API-AGRO (Paris, https://api-agro.eu/, last accessed 19th April 2019) were integrated in the GIS- project. The surface-near geological units are mostly dominated by Quaternary alluvial sedimentation of the River Rhine and River l’Ill. Soil formation pro- cesses and drainage potential are strongly linked to the height of the groundwater level below the sur- face, late Pleistocene and early Holocene loess co- ver, periodic flooding events, and sediment reloca- tions that represent a conglomerate of different cli- matic and geomorphological components (Hage- dorn, Boenigk, 2008; Himmelsbach et al. 2015a; Kempf 2018; 2019a; 2019b; Pfister et al. 2006; Preusser 2008; Preusser et al. 2016; Rentzel et al. 2009). Slope inclination and terrain roughness play a minor role in the study area, although the hydro- logical and geomorphological parameters are subject to natural transport, displacement and sedimentation processes, which are controlled by the gradient. Two landscape models have been calculated from the multivariate environmental datasets. The first samples all information that is supposed to be deci- Fig. 1. Study area and single components of the multivariate environmental model. The Alsace is situ- ated west of the River Rhine, stretching towards the Vosges mountains. Geological data (alluvial deposits) indicate fine-grained material, that lead to clayey-loamy soil conditions with low drainage potential. A high aquifer (processed and interpolated from 327 groundwater stations) and periodic flooding events (processed from Sentinel-1 SAR data from January 2018) lead to locally unfavourable surfaces. Forest coverage, agricultural exploitation, and increasing demand for arable land and infra- structural developments have had a significant impact on the surface over the past 150 years. The mod- ern hydrological network is subject to manifold anthropogenic overprints such as canalization and drainage activities, which reshaped the environment and caused groundwater lowering and erosion. Fables of the past> landscape (re-)constructions and the bias in the data 481 sive for the choice of potential human utilization: adequate drainage potential, aquifer height below 0.5m, non-alluvial geology, very low-flooding vulne- rability, and non-forested areas. The multivariate mo- del generates six suitability classes from 5 (= very high surface suitability, all classes represent excel- lent surface and subsurface conditions) to 0 (= se- vere surface unsuitability, all classes represent se- verely unfavourable surface and subsurface condi- tions). The suitability model visualizes all environ- mental conditions that distinguish potential settle- ment and land-use corridors from areas with un- suitable surface and subsurface conditions based on the evaluation of their qualitative location factors. The second model represents the modern biased surface conditions in the study area. The dominant parameters are deforestation, modern hydrological system, intense modern built-up change, and exten- sive agricultural utilization. The variables create a landscape model with five classes from 0 (= no mo- dern bias) to 4 (= very strong bias). Quantitative analysis of the archaeological record The distribution of archaeological finds in the Alsace is used for the quantitative evaluation of land-use spread and bias through modern infrastructural con- struction activities. The spatial analysis is based on the consistent archaeological database that is pro- vided by the Université de Strasbourg. The project ArkeoGIS is supported by over 170 international in- stitutions and gathers archaeological data from all over the world (arkeogis.org) (Bernard 2019). Origi- nally designed as a local open-source online GIS for the Upper Rhine Valley, the databases hosted by ArkeoGIS now include a vast amount of geospatial data, archaeological sites and environmental maps. Major advantages arise from the large amount of data for each respective archaeological period and continuous updates by Dr Loup Bernard (UMR 7044 ArcHiMedE, Université de Strasbourg). In particular, the chronological differentiation allows one to per- form point pattern analyses that distinguish patterns pertaining to different chronological periods. The database contains a mixture of structured and un- structured data sets that require filtering in an infor- mation system (Gattiglia 2015). The fact that the database consists of both archaeological excavation data and archaeological survey data (including scat- tered and stray finds) poses a particular challenge for the interpretation of the spatial context of the data distribution. In particular survey data are cha- racterized by the specific teleological research foci, the individual interests of the researcher, technical standards, and the site-specific conditions of the se- lected study area (Cowley 2016; van Leusen 1996). For this research the stable versions of the open source software QGIS 2.18.6 and QGIS 3.6.0 (Open Fig. 2. Recent and historical landcover change and land-use in the study area based on various envi- ronmental datasets and remote sensing applications. Multispectral satellite imagery analysis and veg- etation indices (NDVI) reveal massive deforestation processes between 1972 and 2018 (a), extensive crop cultivation (b), and strong built-up change (c). Michael Kempf 482 Source Geospatial Foundation Project, http://qgis.os geo.org) which include GRASS GIS 7.2.0 and GRASS GIS 7.6.0 (Geographic Resources Analysis Support System, http://grass.osgeo.org) were used. The envi- ronmental modelling was supported by spatial sta- tistical analyses conducted in R (R 3.5.1) and R Stu- dio (R Studio 1.2.1335). Point pattern analysis Enrico Crema et al. (2010.1118) described point pat- tern analysis (PPA) as a method that “examines the spatial configuration of point observations across a study area and, potentially, the underlying pro- cess behind its information” (see also Bevan, Co- nolly 2006; Conolly, Lake 2006). The research in the Upper Rhine Valley relies on an archaeological data- set that consists of 10 726 sites that were tested for their clustered behaviour around modern agglom- erations or along linear structures. In combination with Kernel Density Estimates (KDE) and Complete Spatial Randomness tests (CSR), PPA identifies the statistically significant characteristics of a dispersal of points/sites. A short explanation of the most im- portant methods and tests follows. Intensity analysis Intensity analysis, also known as density analysis, is a method that allows one to describe the changing frequencies of observations in the data (Conolly, Lake 2006; Herzog, Yépez 2013). One way to pro- duce intensity estimations is to describe the amount of observations in a geometrical area – usually a re- gular grid. The total amount of observations in each cell can be measured and interpolated from the cells to the entire study area (Herzog, Yépez 2013). The most common interpolation method is Kernel den- sity estimation (KDE), which produces smooth visu- alizations of the point pattern distributions from the core areas and their surroundings (Bonnier et al. 2019; Conolly, Lake 2006). A kernel – which can be visualized as a hill with a particular height, radius, and shape of slope – is placed over each point, and all of the kernels are added together to produce a density map, sometimes called a ‘heat map’. In a GIS, KDE can be processed using the different radii (band- widths) through which the density levels were pro- cessed (Baxter, Beardah 1997; Herzog, Yépez 2013). The radius, however, depends on the subjective re- search question and extent of the study area (Bon- nier et al. 2019; Brigand, Weller 2018; Hughes et al. 2018). Complete Spatial Randomness (CSR) and Ripley’s K-function Spatial point pattern analysis examines the depen- dence between points. The difference with typical point analysis is the inclusion of spatial attributes in a model. The character of the spatial behaviour of point patterns is among the first statistical analyses that are conducted to identify clustering, regular, or dispersed point distribution patterns (Fig. 3). Typi- cally, so-called CSR- tests (Complete Spatial Random- ness) are applied to compare spatial point patterns to complete spatial random processes (Lucio, Caste- lucio de Brito 2004). Oliver Nakoinz and Daniel Knitter (2016) pointed out that CSR-tests allow not only for the detection of random distributions but also of regular point (negative interaction) and clustered point distribu- tions (positive interaction). Points do not behave equally at all scales. At smaller scales, they can show clustered behaviour that gets random or dispersed at larger scales. If the spatial pattern is not clustered, it is either random or regularly dispersed. However, the regular distribution of anthropogenic or ecolo- gical samples is very rare (Haase 1995). One of the most useful statistical approaches to test CSR is Rip- ley’s K-function (Bevan, Conolly 2006; Conolly, Lake 2006) that describes how point patterns are distri- buted over a certain area (Dixon 2002). Ripley’s K defines the radius at which clustered behaviour is established. Broadly speaking, the function counts the number of points within given distances around each point and compares the result to the number of points one would expect within a totally random point distribution. If the number of empirically ob- served points within a certain distance is greater than the number of the simulated random distribu- tion, the empirical point pattern is clustered at that scale. If the number is smaller than the simulation, the distribution is dispersed (Dixon 2002). PPA was conducted in the study area to estimate the spatial behaviour of the archaeological record and the spatial relationship between the record and the modern agglomerations (Fig. 4). First, a grid of 10 x 10km was established across the research area and Fig. 3. Three spatial point patterns: Clustered, regularly dispersed, and random distributions of 49 points. Fables of the past> landscape (re-)constructions and the bias in the data 483 Fig. 4. Point pattern analysis and interpolated density estimates of the site distribution of archaeologi- cal sites and modern agglomeration centroids. (a–1) total count of archaeological sites in a 10 x10km grid; (a–2) modern agglomeration centroids in the same grid. The total number of sites was classified in categories 1–10 (11) with 1 = low number of sites and 10 = high number of sites, and 11 for the out- lier value of 998 sites (a–3, a–4). These reclassified values were assigned to cells in a raster (b–1, b–2) and the differences between both data sets were calculated (b–3). Multilevel b-spline interpolations of these reclassified values (c–1, c–2) and the differences between them were calculated to visualize areas of congruence (c–3, moderate values) and difference (c–3, extreme negative and positive values). Michael Kempf 484 Fig. 5. Density estimates (KDE) of (a) the archaeological sites (r = 5000m, n = 10 726) and (b) the mod- ern agglomeration centroids (r = 10 000m, n = 1913). the total number of archaeological sites and modern agglomeration centroids were calculated for each grid cell. The numbers were reclassified in ranges from 1 to 10 and one outlier 11 (the area of Stras- bourg with 998 archaeological sites in the grid cell). The reclassified values were mapped accordingly (Fig. 4a). From the reclassification, a raster analysis was performed that attaches the number of record- ed sites to every grid cell. The difference between the archaeological and modern raster indicates the high spatial interdependencies of the archaeological sites and the modern agglomerations (low values, –1, 0, 1, white signature in Fig. 4b-3). Areas that are significantly different show high negative or high positive values. From the raster, a multilevel b-spline interpolation was used to produce a density plot with smoothed value ranges (Fig. 4c–1,c–2). Finally, the difference calculation quantifies the spatial rela- tionship estimate of both datasets in the study area. Furthermore, a KDE estimation was performed for both datasets with r=10 000m (Fig. 5b) for the mo- dern agglomeration centroid dataset and r = 5000m for the archaeological sites (Fig. 5a). Thresholds have been calculated to classify the results of the KDE and enhance their visual intelligibility. Both analyses in- dicate spatial interdependencies between the data- sets. However, significant outliers are visible that are caused by extreme values in the point pattern distri- bution. Results and discussion In the Alsace, the bias model reveals a very strong relationship between the spatial distribution of the archaeological record and the modern residential and industrial areas (Fig. 6). To refine the model, the modern agglomeration boundaries were sepa- rated into small residential districts, local industrial areas, and rural complexes. The centroid of every modern built-up complex was calculated (n = 1913) and analysed according to the methods applied to the archaeological database. Most of the sites of both datasets are situated in areas that experienced strong surface transformation. Figure 6 shows the spatial relationship between the distribution of modern residential areas, the archaeological sites, and the accumulative bias surface in the study area. The bias surface was calculated from built-up change, defor- estation, modern arable land-use, and the connec- tion to the modern hydrological network. Five bias classes have been deduced from the accumulative Fables of the past> landscape (re-)constructions and the bias in the data 485 surfaces, which classify the influence of modern use from low to high. The distribution of archaeological sites and modern agglomeration centres modelled on the bias surface enables the estimation of similar spatial behaviour. The distribution patterns of mo- dern agglomerations indicate that a few centres do not show any biased values. This is because modern urban development and new construction sites hard- ly interfere with the historical village centres. The distribution of archaeological finds is similar to that of modern agglomerations. Twenty percent of the archaeological finds show little or no impact from modern land-use. This may be because the archaeo- logical database includes medieval and early mod- ern heritage sites, which are located in the historical centres outside the bias categories. However, 80% of the total archaeological record lies in the biased categories. The datasets have further been analysed using Ripley’s k-function to test CSR (Fig. 6e,f). The results reveal significant clustering, and a random distribution can be excluded. This supports the argu- ment for a strong spatial relationship between the two site distributions. These estimates indicate intensive location relation- ships between modern development, geomorpholo- gy, vegetation cover, land use, and the distribution of archaeological sites. Due to the similar spatial behaviour of settlement centres and archaeological sites, the hypothesis is pursued that modern con- struction activity is the decisive factor in the per- ception of archaeological concentration areas. For this reason, Thiessen/Voronoi polygons were calcu- lated and analysed for their size and the spatial re- lationships with the archaeological record. The poly- gons are not randomly distributed over the area and their size is strongly linked to the highest modern built-up density and the most intensive construction activity (for example, in the agglomerations of Stras- bourg). Intensively restructured areas represent a small network of polygons, while rural areas are characterized by larger polygons. The archaeologi- Fig. 6. Bias model from built-up change, deforestation, and the modern hydrological network in the Alsace (a). Bright areas show low bias intensity, dark areas high bias intensity. Modern agglomera- tions (green) and the archaeological record (red) are modelled to estimate the bias value of each site (b, c). Both site distributions show clustered spatial behaviour and are not randomly dispersed (e, f). Michael Kempf 486 Fig. 7. Calculated Thiessen polygons from modern rural and urban agglomeration centroids (a). The size varies from 0.0007km2 to 35km2 in the study area. Small polygons indicate high population density (high modern residential area density). The archaeological sites are homogeneously distributed in the poly- gons with only a few outliers caused by archaeological site concentrations in the area of Strasbourg (b, c). The polygons show clustered spatial behaviour and only a few polygons reach up to more than 15km2 (d). Most of the sites are situated in small urban and rural agglomerations. There is no signif- icant correlation between increasing polygon size and increasing number of archaeological sites. cal sites are homogeneously distributed in the poly- gons. Only a few polygons show extreme values (the agglomerations of Strasbourg). Most of the sites are situated in small urban and rural agglomerations. There is no significant correlation between increas- ing polygon size and an increasing number of ar- chaeological sites. The observations from the K-func- tion are supported by the analyses of the spatial pat- terns of the modern centroid Thiessen/Voronoi poly- gons in relation to the archaeological record. The distribution indicates non-regularly dispersed site distribution (Fig. 7). Furthermore, the land-use potentials of the region were analysed to evaluate continuous site occupa- tion. The multivariate suitability model described above was used as a basis for the spatial analysis of both datasets (Fig. 8). Both site distributions show similar patterns. The modern agglomerations and the archaeological record are distributed within the highest classes of the model (Fig. 8b,c). Ninety-two percent of the modern settlement and built-up cen- troids are located in the two highest suitability cate- gories. The site dispersal decreases significantly with- in the other ranges. A similar signal can be detected in the archaeological record: 91% of the total archae- ological finds lie in the highest two categories, with a sharp decrease in the numbers in the others. An additional distance matrix was calculated to demon- strate the strong spatial relationship between archae- ological sites and modern agglomeration. This re- veals that the closer to an urban or rural agglome- ration an area is, the more archaeological sites can be recorded. The significance is further increased if Fables of the past> landscape (re-)constructions and the bias in the data 487 the distance matrix is based on the residential boun- daries (polygon) instead of the centroids (Fig. 8d,e). The interrelationships are not only visually and spa- tially significant, but also statistically. Modern land- use influences our perception of the distribution of archaeological finds in the landscape. There are se- veral reasons for this: first, it is possible that the study area has experienced continuous utilization and the archaeological sites are located where con- tinuous land-use takes place. This would support the theory of constant settlement and land-use strate- gies, and ignore dynamic environmental and socio- cultural behaviour for several thousand years. Such hypotheses are currently questioned by geologists and Quaternary sedimentologists that are evaluating the palaeochannel shifts and riverbed relocations of the Upper Rhine during the Holocene (Rambeau et al. 2019). The first results indicate strong displace- ments of the Rhine course and its tributaries – over the entire Holocene. According to the authors, large- ly stable conditions of the fluvial system of the river Rhine can only be assumed for the post-Roman peri- od onwards – at least for the investigated parts of the river course. Local settlement continuity can only be assumed for the margins of the higher moun- tain foreland and elevated Mesozoic plateaus in the floodplain. The floodplain of the Holocene anasto- mosing river periodically shifted, and erosion and accumulation processes replaced each other in very dynamic systems that still seem to be unconsidered in landscape archaeology. However, a reasonable question here is whether the suitability model is biased by modern perceptions of the landscape. Entire past landscapes cannot be re- constructed because past cognitive concepts of how landscapes were formed cannot be perceived by mo- dern individuals. Premodern landscapes consist of experiences, traditional values and ideas rather than Fig. 8. Multivariate landscape suitability model (a) composed of soil quality, geological units, low flood vulnerability, high drainage potential and low aquifer (no groundwater discharge). Based on the model, the distribution of modern agglomerations (n = 1913) was analysed. (b) A total of 1161 sites are located in very high and 607 in high suitability classes. The total archaeological record (n = 10 726) shows 5299 sites in very high and 4485 sites in high suitability classes (c). The distance matrix between the archaeological record and the modern agglomerations (d, e) demonstrates that most archaeological sites are located in close proximity to the nearest modern rural or urban agglomera- tion centroids (d). Modelling the distance between the archaeological record and the boundaries of the modern residential polygons further increases the significance (e). Michael Kempf 488 their actual geographical contents (Gramsch 1996). The landscape palimpsests of a variety of cultural human-environment interactions have led to a mas- sive transformation of the earth’s surface, and even- tually to the outlook of the modern world. All archa- eological distribution is finally a modern perception of how we interpret past human behaviour. This is triggered through modern urban agglomerations and the pull-factor of continuously inhabited regions. In- tensive survey activity generates a high archaeolo- gical density in these areas, while adjacent areas show a low data volume due to lower survey inten- sity. Archaeological corridors are created technically and methodically (Armit et al. 2014; van Leusen 1996; van Leusen, Kamermans 2011). The major bias factor is the vicinity to modern built-up areas, and in particular intensive construction activity in the marginal zones of urban agglomerations, exten- sive infrastructure and rail tracks. Furthermore, cul- tural heritage sites in historical centres are important public pillars that acquire increased cultural percep- tions. Well-organized monument preservation man- agement and a high density of excavation compa- nies increases the capability to undertake archaeo- logical surveys and prospections what strengthens public recognition and financial support – in addi- tion to the benefits of potential scientific publication. Conclusion Simple distribution maps of archaeological data are useless. They produce dehumanized patterns in arti- ficial space. The strength of GIS in archaeology is its diversity (Conolly, Lake 2006). The process behind the application of GIS, or digital modelling in gener- al, is not meant to stand opposed to the interpreta- tion of human patterns, but rather to complement and extend the approaches of a comprehensive and modern landscape archaeology (Llobera 2012). Just like in any other science, uncertainties are a funda- mental property of progress and research develop- ment, and archaeological data in particular can eas- ily be confused with absolute data. However, it is the current state of archaeological research that is used to model the spatial behaviour of past soci- eties. The results of the bias and suitability models of the Alsatian Upper Rhine can be used to identify continuously used areas of intense human activity. On the other hand, they can also be used to estimate the impact of modern landcover change on the (mo- dern) archaeological distribution, and thus to engage in methodological source criticism. This paper shows that there are very significant relationships between modern anthropogenic surface modifications and the density of the archaeological record that is perceived by individuals today. Past societies did not leave traces in linear patterns. The perception of cultural heritage is constructed by modern individuals mov- ing in space. The actual archaeological traces were constructed by individuals creating space. That dif- ference can be an additional way to understand past human-environment interactions. The statistical analyses of this article strongly benefit from the discussions with Jan-Eric Schlicht and Oliver Nakoinz (both Kiel University). I am further very gra- teful to Jan Kolář and a second reviewer for their constructive ideas and comments that increased the structure of the paper. ACKNOWLEDGEMENTS Armit I., Swindles G. T., Becker K., Plunkett G., and Bla- auw M. 2014. Rapid climate change did not cause popula- tion collapse at the end of the European Bronze Age. Pro- ceedings of the National Academy of Sciences of the United States of America 111(48): 17045–17049. https://doi.org/10.1073/pnas.1408028111 Baguette M., Blanchet S., Legrand D., Stevens V. M., and Turlure C. 2013. Individual dispersal, landscape connecti- vity and ecological networks. Biological Reviews of the Cambridge Philosophical Society 88(2): 310–326. https://doi.org/10.1111/brv.12000 Baxter M. J., Beardah C. C. 1997. Some Archaeological Ap- plications of Kernel Density Estimates. Journal of Archa- eological Science 24: 347–354. https://doi.org/10.1006/jasc.1996.0119 Berglund B. E. 2003. Human impact and climate changes- synchronous events and a causal link? Quaternary Inter- national 105(1): 7–12. https://doi.org/10.1016/S1040-6182(02)00144-1 Bernard L. 2019. ArkeoGIS. Archéologies numériques 3(1). https://doi.org/10.21494/ISTE.OP.2019.0354 References ∴ Fables of the past> landscape (re-)constructions and the bias in the data 489 Bevan A., Conolly J. 2006. Multiscalar approaches to set- tlement pattern analysis. In G. Lock, B. Molyneaux (eds.), Confronting Scale in Archaeology: Issues of Theory and Practice. Springer. Boston, MA: 217–234. Bonnier A., Finné M., and Weiberg E. 2019. Examining Land-Use through GIS-Based Kernel Density Estimation: A Re-Evaluation of Legacy Data from the Berbati-Limnes Survey. Journal of Field Archaeology 44(2): 70–83. https://doi.org/10.1080/00934690.2019.1570481 Brandt R., Groenewoudt B. J., and Kvamme K. L. 1992. An Experiment in Archaeological Site Location: Modeling in the Nethderlands using GIS Techniques. World Archaeo- logy 24(2): 268–282. https://doi.org/10.1080/00438243.1992.9980207 Brigand R., Weller O. 2018. Neo-Eneolithic settlement pat- tern and salt exploitation in Romanian Moldavia. Journal of Archaeological Science: Reports 17: 68–78. https://doi.org/10.1016/j.jasrep.2017.10.032 Büntgen U. and 11 co-authors. 2011. 2500 years of Euro- pean climate variability and human susceptibility. Science 331(6017): 578–582. doi: 10.1126/science.1197175 Cao H., Zhang H., Wang C., and Zhang B. 2019. Operatio- nal Flood Detection Using Sentinel-1 SAR Data over Large Areas. Water 11(4): 786. https://doi.org/10.3390/w11040786 Conolly J., Lake M. 2006. Geographical information sys- tems in archaeology. Cambridge manuals in archaeology. Cambridge University Press. Cambridge. http://www.loc. gov/catdir/enhancements/fy0665/2006296605-d.html. Cowley D. C. 2016. What Do the Patterns Mean? Archaeo- logical Distributions and Bias in Survey Data. In M. Forte, S. Campana (eds.), Digital methods and Remote Sensing in Archaeology. Springer. Cham: 147–170. Crema E. R., Bevan A., and Lake M. W. 2010. A probabi- listic framework for assessing spatio-temporal point pat- terns in the archaeological record. Journal of Archaeolo- gical Science 37(5): 1118–1130. https://doi.org/10.1016/j.jas.2009.12.012 Dabrowska-Zielinska K., Budzynska M., Tomaszewska M., Malinska A., Gatkowska M., Bartold M., and Malek I. 2016. Assessment of Carbon Flux and Soil Moisture in Wetlands Applying Sentinel-1 Data. Remote Sensing, 8(9): 756. https://doi.org/10.3390/rs8090756 David B., Thomas J. 2010. Landscape Archaeology. In B. David, J. Thomas (eds.), Handbook of landscape archae- ology. Routledge. Abingdon: 27–43. Dixon P. M. 2002. Ripley’s K function. In A. H. El-Shaara- wi, W. W. Piegorsch (eds.), Encyclopedia of environme- trics. Wiley. Chichester: 1796–1803. Doneus M. 2013. Die hinterlassene Landschaft – Pro- spektion und Interpretation in der Landschaftsarchäo- logie. Mitteilungen der Prähistorischen Kommission. Öster- reichische Akademie der Wissenschaften, Philosophisch- Historische Klasse 78. Verlag der Österreichische Akade- mie der Wissenschaften. Wien. Duchne E., Schneider C. 2005. Grapevine and climatic changes: a glance at the situation in Alsace. Agronomy for Sustainable Development 25(1): 93–99. https://doi.org/10.1051/agro:2004057 Fleming A. 2006. Post-processual Landscape Archaeology: a Critique. Cambridge Archaeological Journal 16(3): 267–280. https://doi.org/10.1017/S0959774306000163 Gattiglia G. 2015. Think big about data: Archaeology and the Big Data challenge. Archäologische Information 38: 1–12. https://doi.org/10.11588/ai.2015.1.26155 Giacona F., Martin B., Furst B., Glaser R., Eckert N., Him- melsbach I., and Edelblutte C. 2019. Improving the un- derstanding of flood risk in the Alsatian region by know- ledge capitalization: the ORRION participative observatory. Natural Hazards and Earth System Sciences Discussions 19(8): 1–49. https://doi.org/10.5194/nhess-2018–210 Gibson J. J. 1979. The Ecological Approach to Visual Perception. Houghton Mifflin. Boston, MA. 1986. The ecological approach to visual perception. Erlbaum. Hillsdale, New York. Gillings M. 2007. The Ecsegfalva landscape: affordance and inhabitation. In A. Whittle (ed.), The Early Neolithic on the Great Hungarian Plain: Investigations of the Kö- rös culture site of Ecsegfalva 23, County Békés. AKA- PRINT Nyomdaipari Kft. Budapest: 31–46 2009. Visual Affordance, Landscape, and the Megaliths of Alderney. Oxford Journal of Archaeology 28(4): 335–356. https://doi.org/10.1111/j.1468-0092.2009.00332.x 2012. Landscape Phenomenology, GIS and the Role of Affordance. Journal of Archaeological Method and Theory 19(4): 601–611. https://doi.org/10.1007/s10816-012-9137-4 Glaser R. and 18 co-authors 2010. The variability of Euro- pean floods since AD 1500. Climatic Change 101(1–2): 235–256. https://doi.org/10.1007/s10584-010-9816-7 Michael Kempf 490 Glaser R., Riemann D., Himmelsbach I., Drescher A., Schön- bein J., Martin B., and Vog S. 2012. Analyse historischer Hochwasserereignisse – Ein Beitrag zum Hochwasserrisi- komanagement. Erfahrungsaustausch Betrieb von Hoch- wasserrückhaltebecken in Baden-Württemberg 18. Be- richtsband: 8–17. Gramsch A. 1996. Landscape Archaeology: Of making and seeing. Journal of European Archaeology 4: 19–38. https://doi.org/10.1179/096576696800688060 Groenhuijzen M. R. 2019. Palaeogeographic-Analysis Ap- proaches to Transport and Settlement in the Dutch Part of the Roman Limes. In P. Verhagen, J. Joyce, and M. R. Groenhuijzen (eds.), Finding the Limits of the Limes. Springer International Publishing. Cham: 251–269. https://doi.org/10.1007/978-3-030-04576-0_12 Güimil-Fariña A., Parcero-Oubiña C. 2015. “Dotting the joins”: a non-reconstructive use of Least Cost Paths to ap- proach ancient roads. The case of the Roman roads in the NW Iberian Peninsula. Journal of Archaeological Science 54: 31–44. https://doi.org/10.1016/j.jas.2014.11.030 Gurrutxaga M., Lozano P. J., and del Barrio G. 2010. GIS- based approach for incorporating the connectivity of eco- logical networks into regional planning. Journal for Na- ture Conservation 18(4): 318–326. https://doi.org/10.1016/j.jnc.2010.01.005 Haase P. 1995. Spatial pattern analysis in ecology based on Ripley’s K-function: Introduction and methods of edge correction. Journal of Vegetation Science 6(4): 575– 582. https://www.jstor.org/stable/3236356 Hagedorn E.-M., Boenigk W. 2008. The Pliocene and Qua- ternary sedimentary and fluvial history in the Upper Rhine Graben based on heavy mineral analyses. Netherlands Journal of Geoscience 87(1): 21–32. https://doi.org/10.1017/S001677460002401X Herzog I., Yépez A. 2013. Least-Cost Kernel Density Esti- mation and Interpolation-Based Density Analysis Applied to Survey Data’. In F. Contreras, M. Farjas, and F. J. Mele- ro (eds.), Fusion of cultures. Proceedings of the 38th An- nual Conference on Computer Applications and Quantita- tive Methods in Archaeology, Granada, Spain, April 2010. British Archaeological Reports IS 2494. Archaeopress. Oxford: 367–374. Himmelsbach I., Glaser R., Schoenbein J., Riemann D., and Martin B. 2015a. Flood risk along the upper Rhine since AD 1480. Hydrology and Earth System Sciences Discus- sions 12(1): 177–211. https://doi.org/10.5194/hessd-12-177-2015 2015b. Reconstruction of flood events based on docu- mentary data and transnational flood risk analysis of the Upper Rhine and its French and German tributaries since AD 1480’. Hydrology and Earth System Sciences 19(10): 4149–4164. https://doi.org/10.5194/hess-19-4149-2015 Howey M. C. L. 2011. Multiple pathways across past land- scapes: circuit theory as a complementary geospatial me- thod to least cost path for modeling past movement. Jour- nal of Archaeological Science 38(10): 2523–2535. https://doi.org/10.1016/j.jas.2011.03.024 Howey M. C. L., Brouwer Burg M. 2017. Assessing the state of archaeological GIS research: Unbinding analyses of past landscapes. Journal of Archaeological Science 84: 1–9. https://doi.org/10.1016/j.jas.2017.05.002 Hughes R., Weiberg E., Bonnier A., Finné M., and Kaplan J. O. 2018. Quantifying Land Use in Past Societies from Cultural Practice and Archaeological Data. Land 7(1): 9. https://doi.org/10.3390/land7010009 Ingold T. 1992. Culture and the perception of the envi- ronment. In E. Croll, D. Parkin (eds.), Bush base: forest farm: Culture, environment and development. Rout- ledge, London. New York: 39–55. 2000. The perception of the environment: Essays on livelihood, dwelling and skill. Routledge. London, New York. Jung M. 2018. Das objektepistemologische Potential des Affordanzkonzeptes James Gibsons und seine Bedeutung als Grundlage von ‘Objektbiographien‘. Methodologische Anmerkungen und exemplarische Fallstudie. In M. Hil- gert, K. Hofmann, and H. Simon (eds.), Objektepistemo- logien. Zur Vermessung eines transdisziplinären For- schungsraums. Berlin Studies of the Ancient World 59. Pro Business digital printing. Berlin: 135–178. Kaplan G., Avdan U. 2017. Mapping and monitoring wet- lands using Sentinel-2 satellite imagery. ISPRS Annals of Photogrammetry, Remote Sensing and Spatial Informa- tion Sciences IV–4(W4): 271–277. https://doi.org/10.5194/isprs-annals-IV-4-W4-271-2017 Kempf M. 2018. Migration or landscape fragmentation in Early Medieval eastern France? A case study from Nieder- nai. Journal of Archaeological Science: Reports 21: 593– 605. https://doi.org/10.1016/j.jasrep.2018.08.026 2019a. Paradigm and pragmatism: GIS-based spatial analyses of Roman infrastructure networks and land- use concepts in the Upper Rhine Valley. Geoarchaeolog 74(285): 1–12. https://doi.org/10.1002/gea.21752 Fables of the past> landscape (re-)constructions and the bias in the data 491 2019b. The application of GIS and satellite imagery in archaeological land-use reconstruction: A predictive model? Journal of Archaeological Science: Reports 25: 116–128. https://doi.org/10.1016/j.jasrep.2019.03.035 Kupfer J. A. 2012. Landscape ecology and biogeography. Progress in Physical Geography: Earth and Environ- ment 36(3): 400–420. https://doi.org/10.1177/0309133312439594 Landuyt L., Van Wesemael A., Schumann G. J.-P., Hostache R., Verhoest N., and Vancoillie F. 2019. Flood Mapping Ba- sed on Synthetic Aperture Radar: An Assessment of Estab- lished Approaches. IEEE Transactions on Geoscience and Remote Sensing 57(2): 722–739. https://doi.org/10.1109/TGRS.2018.2860054 Lasaponara R., Masini N. 2006. Identification of archaeo- logical buried remains based on the normalized difference vegetation index (NDVI) from Quickbird satellite data. IEEE Geoscience and Remote Sensing Letters 3(3): 325–328. https://doi.org/10.1109/LGRS.2006.871747 2011. Satellite remote sensing in archaeology: past, present and future perspectives. Journal of Archaeolo- gical Science 38(9): 1995–2002. https://doi.org/10.1016/j.jas.2011.02.002 2013. Satellite Synthetic Aperture Radar in Archaeology and Cultural Landscape: An Overview. Archaeological Prospection 20(2): 71–78. https://doi.org/10.1002/arp.1452 Llobera M. 1996. Exploring the topography of mind: GIS, social space and archaeology. Antiquity 70(269): 612– 622. https://doi.org/10.1017/S0003598X00083745 2001. Building Past Landscape Perception With GIS: Understanding Topographic Prominence. Journal of Archaeological Science 28(9): 1005–1014. https://doi.org/10.1006/jasc.2001.0720 2012. Life on a Pixel: Challenges in the Development of Digital Methods Within an “Interpretive” Landscape Archaeology Framework. Journal of Archaeological Method and Theory 19(4): 495–509. https://doi.org/10.1007/s10816-012-9139-2 Loveland K. A. 1991. Social Affordances and Interaction II: Autism and the Affordances of the Human Environ- ment. Ecological Psychology 3(2): 99–119. https://doi.org/10.1207/s15326969eco0302_3 Lucio P. S., Castelucio de Brito N. L. 2004. Detecting Ran- domness in Spatial Point Patterns: A ‘Stat-Geometrical’ Alternative. Mathematical Geology 36(1): 79–99. https://doi.org/10.1023/B:MATG.0000016231.05785.e4 Malekmohammadi B., Jahanishakib F. 2017. Vulnerability assessment of wetland landscape ecosystem services using driver-pressure-state-impact-response (DPSIR) model. Eco- logical Indicators 82: 293–303. https://doi.org/10.1016/j.ecolind.2017.06.060 Masini N., Soldovieri F. (eds.) 2017. Sensing the past: From artifact to historical site. Geotechnologies and the environment, volume 16. Springer. Cham. Meier T. 2009. Umweltarchäologie – Landschaftsarchäolo- gie. In S. Brather (ed.), Historia archaeologica: Fest- schrift für Heiko Steuer. Ergänzungsbände zum Reallexi- kon der germanischen Altertumskunde 70. de Gruyter. Berlin: 697–734. 2017. Potenziale und Risiken der Umweltarchäologie. In S. Brather, J. Dendorfer (eds.), Grenzen, Räume und Identitäten: Der Oberrhein und seine Nachbarregio- nen von der Antike bis zum Hochmittelalter. Jan Thor- becke Verlag. Ostfildern: 13–53. Mleczko M., Mróz M. 2018. Wetland Mapping Using SAR Data from the Sentinel-1A and TanDEM-X Missions: A Comparative Study in the Biebrza Floodplain (Poland). Remote Sensing 10(2): 78. https://doi.org/10.3390/rs10010078 Morrison K. 2013. Mapping Subsurface Archaeology with SAR. Archaeological Prospection 20(2): 149–160. https://doi.org/10.1002/arp.1445 Muthers S., Laschewski G., and Matzarakis A. 2017. The Summers 2003 and 2015 in South-West Germany: Heat Waves and Heat-Related Mortality in the Context of Cli- mate Change. Atmosphere 8(224): 1–13. https://doi.org/10.3390/atmos8110224 Pfister C., Weingartner R., and Luterbacher J. 2006. Hydro- logical winter droughts over the last 450 years in the Up- per Rhine basin: a methodological approach. Hydrologi- cal Sciences Journal 51(5): 966–985. https://doi.org/10.1623/hysj.51.5.966 Preusser F. 2008. Characterisation and evolution of the River Rhine system. Netherlands Journal of Geoscience 87(1): 7–19. https://doi.org/10.1017/S0016774600024008 Preusser F., May J.-H., Eschbach D., Trauerstein M., and Schmitt L. 2016. Infrared stimulated luminescence dating of 19th century fluvial deposits from the upper Rhine Ri- ver. Geochronometria 43(1): 131–142. https://doi.org/10.1515/geochr-2015-0045 Rambeau C., Schmitt L., Chapkanski S., Houssier J., Ertlen D., and Schneider N. 2019. Fluvial dynamics in the Upper Rhine Graben (NE France) for the past ca. 12.000 years – Michael Kempf 492 input from palaeochannel dating and provenance stud- ies. In 17ème Congrès Français de Sédimentologie, Bea- uvais, 22 – 24 Octobre 2019, Livre des résumés. Publica- tion Association des Sedimentologistes Français 81. Paris. Rentzel P., Preusser F., Pümpin C., and Wolf J.-J. 2009. Loess and palaeosols on the High Terrace at Sierentz (France), and implications for the chronology of terrace formation in the Upper Rhine Graben. Swiss Journal of Geosciences 102(3): 387–401. https://doi.org/10.1007/s00015-009-1338-9 Schaich H., Bieling C., and Plieninger T. 2010. Linking Ecosystem Services with Cultural Landscape Research. GAIA – Ecological Perspectives on Science and Society 19(4): 269–277. https://doi.org/10.14512/gaia.19.4.9 Shen X., Wang D. Mao K., Anagnostou E., and Hong Y. 2019. Inundation Extent Mapping by Synthetic Aperture Radar: A Review. Remote Sensing 11(7): 879. https://doi.org/10.3390/rs11070879 Stratoulias D., Balzter H., Zlinszky A., and Tóth V. R. 2018. A comparison of airborne hyperspectral-based classifica- tions of emergent wetland vegetation at Lake Balaton, Hungary. International Journal of Remote Sensing 39 (17): 5689–5715. https://doi.org/10.1080/01431161.2018.1466081 Toohey M., Krüger K., Sigl M., Stordal F., and Svensen H. 2016. Climatic and societal impacts of a volcanic double event at the dawn of the Middle Ages. Climatic Change 136(3–4): 401–412. https://doi.org/10.1007/s10584-016-1648-7 van Dinter M. 2013. The Roman Limes in the Netherlands: how a delta landscape determined the location of the mi- litary structures. Netherlands Journal of Geoscience 92 (1): 11–32. https://doi.org/10.1017/S0016774600000251 van Leusen M., Kamermans H. (eds.) 2011. Predictive Mo- delling for Archaeological Heritage Management: A re- search agenda. Rijksdienst voor het Oudheidkundig Bo- demonderzoek. Amersfoort. van Leusen P. M. 1996. Unbiasing the Archaeological Re- cord. Archeologia e Calcolatori 7: 129–136. Verhagen J. W. H. P. 2007. Case studies in archaeologi- cal predictive modelling. Archaeological studies Leiden University 14. Leiden University Press. Leiden: Verhagen P., Kamermans H., van Leusen M., Deeben J., Hallewas D. P., and Zoetbrood P. 2010. First Thoughts on the Incorporation of Cultural Variables into Predictive Modelling. In F. Niccolucci, S. Hermon (eds.), Beyond te Artifact: Digital Interpretation of the Past. Proceedings of CAA2004. Prato 13–17 April 2004. Archaeolingua. Bu- dapest: 307–311. Verhagen P. 2018. Predictive Modeling. In S. L. López Va- rela (ed.), The encyclopaedia of archaeological sciences. Wiley-Blackwell. Malden, MA: 1–3. Weaverdyck E. J. S. 2019. The Role of Forts in the Local Market System in the Lower Rhine: Towards a Method of Multiple Hypothesis Testing Through Comparative Model- ling. In P. Verhagen, J. Joyce, and M. R. Groenhuijzen (eds.), Finding the Limits of the Limes. Springer Interna- tional Publishing. Cham: 165–190. Webster D. S. 1999. The concept of affordance and GIS: a note on Llobera (1996). Antiquity 73(282): 915–917. https://doi.org/10.1017/S0003598X00065698