The use of LiDAR in reconstructing the pre‐World War II landscapes of abandoned mountain villages in southern Poland

The cessation of most human activities resulting from post‐World War II expulsions and forced displacements in Central Europe triggered massive land cover transformation in mountainous areas. However, many pre‐War traces of past landscapes have survived—imprinted in microtopography—in permanently abandoned villages. Currently, they constitute unique cultural heritage of communities no longer in existence. Our main goal was therefore to reconstruct a lost cultural landscape of mountain villages abandoned after World War II (WWII). The case study area comprised three such villages located in southern Poland, two in the Carpathians and one in the Sudetes. We used the national airborne light detection and ranging (LiDAR) dataset combined with archival cadastral maps and field survey to detect man‐made microtopographic features related to past boundaries, road network, agriculture and buildings and to interpret them in the landscape context. We demonstrated that the pre‐War human footprint left in relief was shaped largely by past landownership divisions, land use and environmental constraints (related to lithology, soils and topography). Our secondary goal was to assess the value and application opportunities of LiDAR in reconstructing past landscapes. We showed that 38–70% of non‐natural parcel boundaries and 65–79% of roads marked on mid‐19th‐century cadastral maps are still detectable using LiDAR. Therefore, we argue that the past landscape pattern, originating in late Middle Ages and subsequently transformed prior to WWII, remains well preserved in the relief and that LiDAR is an effective tool to reconstruct a past landscape of mountain villages abandoned after WWII. We also confirmed that customized LiDAR visualizations are more informative than ready‐to‐use shaded digital elevation models (DEMs), in particular when integrated with cadastral and field‐based data. We conclude that the greatest advantage of LiDAR is the capacity to provide a landscape context for isolated traces of past human activity, allowing for the reconstruction of entire spatial patterns and interrelationships developed by past societies.


| INTRODUCTION
The reconstruction of past landscapes provides a basis for understanding and predicting human-environment interactions in wider spatial contexts (Bürgi et al., 2017;MacDonald et al., 2000;Plieninger et al., 2016). According to European Landscape Convention, landscapes require understanding, protection and sustainable management as they constitute ubiquitous and primary context of shared heritage (Council of Europe, 2000;Millican et al., 2017). Because present-day landscapes are a legacy of past land use activities, the recognition of historical landscape patterns is a matter of both scientific and policy interest Marcucci, 2000;Stabbetorp et al., 2007).
However, as settlement history is studied and past landscapes in various ways reconstructed, the challenge is to uncover scattered traces of past land use and to tie them together. Successive episodes of human activity leave marks in the landscape that are superimposed one upon the other (Bailey, 2007;Mlekuž, 2013a). Therefore, a complex spatial pattern is created down the centuries, resembling a cumulative 'landscape palimpsest', which is often hard to read out and interpret (Marcucci, 2000;Palang et al., 2006), even more so where proper context is lacking (Johnson & Ouimet, 2018).
A fairly accurate and precise reconstruction of the historical landscape can be obtained by analysing multitemporal cartographic sources, especially when georeferenced, digitized and analysed using geographic information system (GIS) tools (Kienast, 1993). Maps based on mathematical foundations and geodetic measurements have been available since the turn of the 18th and 19th centuries. Archival cadastral maps and associated protocols are among the most valuable sources of spatial data. They have been commonly used for the detailed analyses of landscape transformations (Oláh & Boltiziar, 2009;Petek & Urbanc, 2004), delimitation of parcels (Forejt et al., 2018), and reconstructions of past tillage patterns (Domaas, 2007) and terrace fields (Rendu et al., 2015).
Since the early 1930s, aerial photographs have been systematically collected (Pinto et al., 2019), whereas civilian and commercial satellite images with increasingly higher resolution are available since the 1970s (Belward & Skøien, 2015). Both of these data sources have been extensively used to approach past cultural landscapes in mountainous regions (Jabs-Soboci nska et al., 2021;Malek et al., 2014;Millican et al., 2017).
Available spatial data have usually been combined with noncartographic sources, for example, census data , livestock tax registers (Dahlström, 2008), forestry inventories (Saito et al., 2007), national records of sites and monuments (Millican et al., 2017), ground-based repeat photography (Dethier et al., 2018;Kaim, 2017) and also memories of the local people .
An invaluable complement to the proper identification of remnants of historical landscapes is provided by detailed geomorphic (Kirchner et al., 2020;Latocha, 2015b) and botanical (Latocha et al., 2019;Majewska, 2019) field-based data and mapping (Wolski, 2007). By integrating different data, a more accurate picture of the entire spatial patterns can be obtained (Pelorosso et al., 2009;Petit & Lambin, 2002).
However, old maps and aerial and satellite images have clear limitations when it comes to the detection of relic man-made landscape features. They relate to scale and level of generalization (maps), resolution (satellite images) and lighting conditions (aerial images) (Devereux et al., 2005;Latocha, 2015a). Beyond that, major problems arise from the specific nature of passive photogrammetry-as forest canopies or dense scrub limit the visibility of relief. It is because passive sensors measure only the energy available naturally (Manley, 2016).
Currently, one of the most effective tools to identify and map relics of past landscapes over extensive areas (not only in isolated sites), which largely devoid of the limitations of passive photogrammetry, is light detection and ranging (LiDAR)-an active remote sensing technique (Sittler et al., 2007). It provides its own source of energy to illuminate and detect targets. Operational versions of commercial fullwaveform LiDAR systems, in which a theoretically infinite number of echoes can be identified, have been available since 2004 (Mallet & Bretar, 2009). However, not LiDAR data alone, but when combined with other data sources, seem to offer the most promising research prospects (Banaszek, 2020;Buján et al., 2013;Manley, 2016). Among the most valuable auxiliary data are archival cadastral maps (Bailly et al., 2008;Bozek et al., 2017) and other cartographic sources relating to historical ownership (Johnson & Ouimet, 2014) and, last but not least, those derived from field prospection (Horňák & Zachar, 2017;Randall, 2014).
Over the last decade, the most spectacular LiDAR-based reconstructions have been associated with UNESCO World Heritage Sites (Masini & Lasaponara, 2013;Megarry et al., 2016). One of the pioneering studies was conducted at Stonehenge, where new archaeological sites, hitherto concealed beneath vegetation, were revealed (Bewley et al., 2005). Yet the most extensive research of this kind has concerned pre-Columbian human settlement (especially the Maya civilization). The application of LiDAR made it possible to trace the spatiotemporal evolution of the cityscape with complex and highly structured systems of water management (canals and causeways), agriculture (terraces) and defence in Belize (Chase et al., 2014(Chase et al., , 2011, Guatemala (Canuto et al., 2018) and Honduras (Fisher et al., 2016), as well as outside Mesoamerica, for example, in Lower Amazon in Brazil (Stenborg et al., 2018). Similar research has also been carried out in Cambodia, where the capitals of Khmer Empire at Angkor-the largest settlement complex of the preindustrial world-were located (Chevance et al., 2019;Evans et al., 2013). Furthermore, in the Caribbean, it proved possible to reveal historic-era landscapes impacted by volcanic disaster, so far concealed in Neotropical forest (Opitz et al., 2015).
Traces of past human activity are reflected particularly well in the relief of areas with complex natural topography-in the uplands and mountains. Such areas are usually highly forested now and often difficult to penetrate from the ground; therefore, airborne LiDAR is seen as the technique of choice to document the cultural remains in landscapes of this type. LiDAR has been used in the reconstruction of mediaeval fortified settlements in the Apennines in Italy (Masini et al., 2018), German field fortifications from World War II (WWII) in the Polish Carpathians (Jucha et al., 2020), and historical reservoirs in the mountains of Moldova (M arg arint et al., 2021) and Czechia (Langhammer et al., 2018). On the other hand, the technique has also been used to map hollow ways and barrow cemeteries in selected sites in Slovenia (Kokalj & Hesse, 2017), as well as earthworks in the Harz Mountains in Germany (Swieder, 2021).
Apart from discovering the hidden ancient sites or structures, LiDAR technology has also offered unique possibilities to add spatial, landscape context to historical reconstructions. However, despite the well-established position of LiDAR as a tool in archaeological studies, LiDAR-based reconstructions still rarely capture entire landscape patterns (Affek, 2016;Johnson & Ouimet, 2014;van der Schriek & Beex, 2017). Research has been usually focused on selected types of relic man-made features, for example, linear: hollow ways (Kirchner et al., 2020); agricultural terraces (Tarolli et al., 2014); ditch networks (Bailly et al., 2008); point: ancient shell mounds (Randall, 2014); protohistoric semi-subterranean caches (Krasinski et al., 2016); and pattern based: ridge-and-furrow topography (Sittler et al., 2007). Among the very few examples of LiDAR use in landscape-scale archaeological studies are the large-scale mapping programmes mounted in Germany (Hesse, 2013) and also in Scotland (Historic Environment Scotland- Banaszek et al., 2018) and England (Historic England Aerial Investigation and Mapping-Evans, 2019).
An area that was once populated, and permanently abandoned at some point, is perfect for the LiDAR-based reconstruction of former landscape patterns. The relic man-made features may indeed persist in a landscape provided that they are sufficiently resilient to processes of natural decay. However, it is not the fact of abandonment as such, but rather later land use, that shapes the survival pattern of earlier landscape features, with either good or poor conditions offered for the effort to reconstruct past landscapes. Mountain villages permanently abandoned after World War II in Central and Eastern Europe (see, e.g., Kučera & Chromy, 2012;Latocha, 2012;Mares et al., 2013;Palang et al., 2006) provide a unique opportunity for contextual reconstruction of earthworks characteristic for the pre-War rural landscape. Moreover, as pre-War communities no longer exist in the areas concerned, traces of past activities imprinted in the topography down the centuries now constitute the only in situ heritage of former inhabitants, worth recognizing and protecting (Affek, 2016).
The main goal of our study was therefore to reconstruct the lost landscape of abandoned mountain villages in southern Poland using national airborne LiDAR dataset combined with archival and fieldbased data sources. Our specific objectives were, first, to characterize the landscape pattern of selected villages before abandonment and, second, to assess the value and application opportunities of LiDAR in reconstructing the past landscapes of such villages. Our case study samples a phenomenon observable throughout Central and Eastern Europe, that is, permanent abandonment of settlement in the aftermath of widespread post-WWII displacements of population. The case study approach applied allowed for an extraction of more generalized rules shaping past landscapes in now-deserted mountainous or upland areas, while helping augment existing knowledge on particular types of landscape features, to the benefit of both our understanding of mountain regions in Central and Eastern Europe, and future comparative studies and syntheses.

| STUDY AREA
Expulsions and forced displacements during and after World War II took place across most of Central and Eastern Europe. They were a consequence of border shifts and a policy of ethnically homogeneous nation states (Prausser & Rees, 2004;Ther & Siljak, 2001). As a result, vast areas of mountain borderlands least suitable for agriculture became permanently depopulated and abandoned (Bičík & Štěpánek, 1994;Kučera & Chromy, 2012;Soja, 2012). Three rural mountain regions located within the present borders of Poland belong to those depopulated to the greatest extent: the Śnieżnik Massif (in the Sudetes), the Przemy sl Foothills and the Bieszczady Mountains (both in the Carpathians) (Affek et al., 2020;Latocha, 2012;Wolski, 2007). The land of deserted villages usually underwent nationalization and afforestation, though natural secondary succession also ensued. This ensures a situation in which extensive areas are now covered by postagricultural forests with many man-made topographic forms hidden beneath the tree canopy (Affek, 2016;Latocha, 2015b). Concomitantly, the area of open landscape has shrunk considerably, remaining mostly in places where large-scale grazing was brought in during the post-War period (Wolski, 2007).
For the purposes of our study, each of the regions referred to above is represented by the single fully deserted village considered to have been least impacted by humans since the time of abandonment, that is, over the last 70 years. The villages meeting these criteria are Rog ożka, Borysławka and Cary nskie (Table 1 and Figure 1). Soon after the War, Borysławka was subjected to almost total afforestation. In contrast, Cary nskie and Rog ożka have retained substantial shares of open land, mostly grassland, until present. Equally, the ground point densities measured at each study site  and Cary nskie-3.00) do justify use of the higher than default DEM resolution. The increased resolution performed much better, particularly when it came to the detection of discrete linear features such as road ruts and ploughing traces ( Figure 2).

| LiDAR data
In addition to resolution, the interpolation algorithm used had a substantial effect on the accuracy and scope of image interpretation (Kokalj & Hesse, 2017). We tested two popular interpolation methods called inverse distance weighting (IDW) and triangulated irregular network (TIN). When we applied IDW, the image was blurred, the details located within the flat bottoms of valleys were undistinguishable, the road ruts and ploughing traces become barely visible, and interpolation artefacts appeared in areas with lower point densities. In contrast, the TIN interpolation, which we finally used, performed much better ( Figure 3). The conversion from point cloud to DEM (only 'ground' class points, resolution 0.5 m, TIN interpolation, natural neighbour sampling and no thinning) was then conducted using LP360 Advanced software.
Next, we tested various visualization techniques, for example, shaded relief, openness, local dominance, sky-view factor (SVF) and composite images that were the combinations of the above (Kokalj & Hesse, 2017). Ultimately, we decided to use SVF because the comparative analysis carried out revealed SVF as the most informative technique and easiest to interpret in the context of our study areas and research purposes. Furthermore, SVF is recommended for areas of moderate to steep topography, given the way in which this highlights surface depressions (e.g., hollow ways and ploughing patterns) and features on slopes (e.g., agricultural terraces) (Kokalj & Hesse, 2017).
Based on diffuse illumination, SVF is a geophysical parameter that represents the portion of the sky visible from a certain point. In  practice, points on a ridge are brighter than those at the bottom of a valley because larger part of the sky can be seen from there . Our SVF visualization (16 directions, 5-m range, without noise reduction) was generated using the Relief Visualization Toolbox

| Cadastral archival maps
To reconstruct relic cultural landscape patterns in the three villages selected for study, we used the following high-resolution spatial data as reference: (1) 1:2880 Austrian Empire cadastral maps dating back to 1852-1854 and (2) 1:2500 Prussian Empire cadastral maps from 1863. As both map series were developed for fiscal (land tax) and judicial purposes, they pay much attention to details relating to ownership structure and land use, and hence, the boundaries of individual plots, buildings and tax-free land (e.g., roads) were marked in great detail.
Scans of the required maps were obtained from the Polish State Archives in Przemy sl, Rzesz ow and Wrocław.
Processing of spatial data comprised two main stages: georeferencing and vectorization. We georeferenced all sheets in line with the procedure described by Affek (2015). In the case of the Austrian maps, we used the map datum and projection parameters for the Second Military Survey of Galicia and Bukovina (Affek, 2013), whereas for the Prussian maps, we used bilinear interpolation based on 30-40 ground control points per sheet.
Then, we digitized manually selected information from the scanned maps (see Section 3.4), in line with the map legends (where possible verified with original mapping instructions and definitions of map symbols).

| Field work
We had conducted detailed field survey at all the selected test sites as part of our prior research (Affek, 2016;Latocha, 2015b;Wolski, 2007). Our non-invasive field prospection encompassed, among others: (1)

| Data integration and analysis
Scanned archival cadastral maps and LiDAR-derived DEM visualizations (SVF), along with vector layers generated on their basis, were brought to a common coordinate system (Poland CS92; EPSG:2180).
The project was then further augmented by vector layers with data collected in the field, together with geolocalized photographs. Cadastral data and field survey served as a support in identification of earthworks with pre-WWII origin reflected in LiDAR-based DEM visualizations (in relation to dating, and the determination of origins and functions). We characterized the physical attributes (quantity, course, shape, visibility, etc.) of village boundaries, strips of fields, agricultural terraces, ploughing patterns, mounds, roads, remnants of buildings and mining-related earthworks ( Table 3).
As our second objective was to assess the effectiveness/ usefulness of LiDAR in reconstructing the past landscapes of the abandoned villages, we performed comparative visibility analysis. We sought to determine how much of the landscape pattern recorded on the mid-19th-century cadastral maps is still detectable by LiDAR in each studied village. To this end, we analysed three anthropogenic landscape elements, that is, the boundaries of agricultural parcels (excluding natural boundaries), cadastral roads and the remnants of farmsteads (Table 4). We generated three vector layers showing the respective elements as they were presented on cadastral maps.
Then, we compared them with SVF DEM visualization and assigned the binary visibility category (visible/invisible) to each line section (roads and boundaries) and polygon object (farmsteads). Those sections of old roads that were rebuilt completely after WWII (hardened, profiled, widened, with corrected course and geometry) were not considered part of past landscape and were therefore classified as 'invisible' (covered by a new layer). In the case of remnants of farmsteads, we considered these visible even when it was possible to detect only a minor man-made disturbance to the natural relief indicative of historical settlement (e.g., a cellar, a well and an unambiguous settlement terrace).

| The past rural landscape in abandoned mountain villages as reflected in LiDAR data
The past landscape of the three mountain villages abandoned after World War II was still reflected in the topography ( Figure 4) and therefore well visible in LiDAR data. The field survey and the analysis of archival cadastral maps allowed for the interpretation of microtopographic features reflected in LiDAR derivatives and for the reconstruction of the pre-WWII landscape (Figures 5 and 6). We described quantitatively several man-made landscape features (earthworks) detected by LiDAR (Table 3) and grouped them into the following four major categories: boundaries, the road network, agriculture and buildings.

| Boundaries
The various types of old boundaries clearly visible on LiDAR-based DEM visualizations include village administrative boundaries, ownership boundaries and land use boundaries. It was possible to track the old village boundary of Rog ożka at its complete length and to a large extent also those of Borysławka and Cary nskie. They either ran along natural landforms (ridges and streams-in that case, additional cartographic materials were needed for boundary identification), or were marked by man by way of a stone wall (in Rog ożka most often along the road), an escarpment or a ditch (in Borysławka). Administrative boundaries also ran partially along old tracks (especially ridge paths used by cattle herders at Cary nskie). In Borysławka, village tripoints were additionally marked using groups of three earth mounds.
The former ownership boundaries dating back to the colonization period (14-16th centuries) were also clearly visible in LiDAR derivatives. Field strips perpendicular to the main axis of settlements were only slightly adapted to the terrain and were bordered by escarpments (up to 3 m high in Borysławka and Rog ożka) or by stone walls and mounds (in Rog ożka and Cary nskie). They also in part ran along secondary streams or rock outcrops (in Rog ożka). In some places, the only evidence of the ownership boundary was a field road or preserved diverse patterns of ploughing between neighbouring pre-War parcels. Originally

| Agriculture
The most characteristic elements of relief related to mountain agriculture are the embankments of agricultural terraces. These earthworks perpendicular to the slope were located at field margins, and they made the ploughing easier and limited erosion. Some coincided with ownership boundaries, whereas others were perpendicular to strips of land and located within one parcel, depending on the topography.
Ploughing pattern was another agriculture-related pre-War feature visible in LiDAR derivatives, but only in Borysławka and Cary nskie.
They are best preserved/most visible in former arable fields located further away from the valley bottom, which were completely abandoned and underwent succession processes directly after displacements, and now are covered by mature forest. In Borysławka, we There are also minor traces of former elements of dwellings, which can be detected (at least partially) by LiDAR. Examples include cellars, wells and earth pits (e.g., for storing potatoes). They were found in all the analysed areas. We also identified free-standing

| The effectiveness of LiDAR in detecting past landscape features of abandoned villages
The capacity of LiDAR derivatives to detect past landscape features marked on mid-19th-century cadastral maps varied among features and locations (Table 4). We noted the highest proportion of visible features in regard to old roads (73% on average) and considerably lower in regard to farmsteads and the boundaries of agricultural parcels (52% on average). Borysławka stands out in terms of the visibility of farmstead remnants (71%), whereas Rog ożka for the high visibility of its old field boundaries (70%).
We compared the usefulness of LiDAR with the two other data sources (archival cadastral maps and field survey), when it came to the detection of past landscape features in the villages studied (  In turn, the differences between the investigated sites can be explained mainly by reference to local environmental constraints, involving topography, elevation and lithology. The very shallow and stony soils developed in crystalline bedrock (at Rog ożka) contributed to a far greater number of stone embankments located along roads and field boundaries, as well as to sizes of clearance cairns larger than those at the sites in SE Poland, where thicker soil overlays the sedimentary rocks (Carpathian flysch in Borysławka and Cary nskie). This fact ensured that farmers in the latter villages were less affected than their counterparts in the Sudetes by the need to remove stones to facilitate or even allow for ploughing.
In turn, the much greater density of agricultural terraces in the Carpathian villages can be explained by the need to make greater efforts to limit surface wash from highly erodible flysch soils. In general, such terraces are among the commonest features of past agricultural landscapes detectable using LiDAR derivatives in the uplands and mountains (Chase et al., 2011;Crow, 2009;Migo n & Latocha, 2018). The flysch structure also favoured the formation of hollow ways. They proved to be much more common and deeper at the two sites in SE Poland than in the Sudetes where development was much hampered by the hard, crystalline bedrock or coarsegrained weathering slope covers.
It also emerges that the old ploughing pattern that might be very persistent in an abandoned landscape (Domaas, 2007;Latocha, 2015b) is actually visible solely where there is sedimentary bedrock. However, we also observed differences between Borysławka and Cary nskie, and they may be attributed to both topography and altitude. Many more stony features (mounds, walls etc.) and agricultural terraces were recorded in the mountainous locations and on steep slopes (at Cary nskie) than in the foothills (at Borysławka). In this aspect, Cary nskie is more similar to Rog ożka, despite substantial differences in lithology. Such differences between LiDAR-detectable features in areas with more and less complex topography resemble those in other regions, regardless of the climatic zone (Canuto et al., 2018;Millican et al., 2017). This shows how knowledge of the local environmental conditions is vital for proper interpretation of LiDAR data and for comparative analysis between sites as regards the potential to preserve (or not) man-made features from the past. Moreover, the wider environmental context provides a wider landscape perspective from which to interpret the isolated traces of past human activities.
Our results can be related to observations made by other

| Added value and limitations of LiDAR-based reconstruction of past landscapes
With the advent of LiDAR, it became possible to reveal, at the landscape scale, traces of former land use hidden beneath vegetation (Mlekuž, 2013b). Thanks to LiDAR, we were able to contextualize single historical features (e.g., farmstead foundations, clearance cairns and hollow ways) and reconstruct entire cultural landscapes well reflected in the relief of abandoned villages. Images from a bird's-eye view helped perceive and interpret organized wholes instead of collections of parts, in line with the Gestalt theory of perception (Farina, 2006). They offered deeper insight into the process by which a landscape could be understood. LiDAR allowed us to reconstruct linear objects with even small, 10-to 20-cm height differences and major discontinuities, for example, old roads and field boundaries now hidden beneath tree canopy. This proved to be particularly true in the case of old ploughing patterns in postagricultural forests, which were extremely hard to detect and map without LiDAR. By recording those patterns, laser scanning offered a unique chance to distinguish between ancient and postagricultural forests and to reconstruct past field mosaics and local agricultural practices.
As However, the most important disruptive factor, especially when detecting subtle artefacts, is site-specific vegetation cover (Guan et al., 2014). When earthworks are covered by low and dense vegetation, there is no possibility of obtaining an accurate shape of the Earth's surface due to scanner limitations manifested in the so-called multitarget resolution (MTR) (Ullrich & Pfennigbauer, 2011). In such cases, even if some of the pulses reach the ground, the nominal ranging accuracy of 2 cm (for single returns) increases up to 60 cm (Wagner et al., 2006). As a result, echoes separated by shorter distances within the same laser shot cannot be distinguished. Consequently, the measured range can only be estimated, and the resulting image is blurred (Di Salvo & Lo Brutto, 2014). This is particularly evident where vegetation porosity is very poor, for example, in the circumstances of a dense understorey with brambles and bracken fronds (Crow, 2009;Doneus et al., 2008), or else where there is thick blanketing moss (Krasinski et al., 2016). Therefore, we argue that  (Affek, 2014;Norstedt et al., 2020). In our case, customizing the entire process of generating final DEM visualizations (selection of points, DEM resolution and visualization technique), instead of working on ready-to-use shaded DEM, allowed for the detection of far more man-made microtopographic features (e.g., road ruts and subtle old ploughing traces) and for dating them more accurately. The presence of road ruts, apart from the U-shape cross-section of hollow ways (Kirchner et al., 2020), was an important indication of current use of roads, so a possibility of detecting them improved the dating of road traces substantially.
On the other hand, the most challenging was the dating of forest roads and minor field roads, as they were not marked on old cadastral maps. Some such roads were used after the War by foresters, but they may as well have been created in the pre-or inter-War periods, which was also noticed by Johnson and Ouimet (2014). Sometimes overgrown, nonhollow dirt roads (especially within a flat valley bottom) simply cannot be fully discriminated from bare earth (Buján et al., 2013), though patterns of artefacts (i.e., sequences of pointless or blurred areas) can also be a hint (Jucha et al., 2020). Still, in many cases, it was possible to distinguish between pre-and post-WWII forest roads, mainly by analysing their course against the landscape pattern. Old forest roads fit much better into the historical landscape pattern and usually lead along ridges, whereas post-War forest roads often cross the pre-War mosaic of fields, are less steep and almost perpendicular to the slope and form a specific network in line with post-War forestry guidelines (Affek et al., 2017). Moreover, modern mechanized skidding with heavy logging equipment leaves visually different traces compared with the historical use of draught horses (Affek et al., 2017), and the 0.5-m resolution DEM allowed us to detect such indicative microtopographic details as the shape of the ruts and overall road profile. But even when all these premises are taken into account, some level of uncertainty in dating remains.
LiDAR has been confirmed as an extremely powerful reconnaissance tool. However, LiDAR alone, and despite optimization efforts, does not suffice for the reliable dating of past landscape features and for clear distinguishing between pre-and post-WWII origins. Joint analysis with archival cadastral maps did much to facilitate the identification of old parcel boundaries, public roads and farmsteads. In turn, field survey contributed to the verification of small concave and convex objects (e.g., clearance cairns, wells and cellars) not marked on any available source of spatial data, including cadastral maps.
In general, LiDAR-based identification is more problematic for point features than linear features (Crow, 2009), though the interpretation of the latter is also not obvious at times, where pedestrian survey data are lacking (Johnson & Ouimet, 2018;Quintus et al., 2017;Risbøl et al., 2013). Firstly, point features can be easily obscured by dense overlying vegetation or omitted where point density is too low.
Moreover, field survey was necessary to create an image interpretation key that linked real landscape features with their reflection on LiDAR-based visualizations (Randall, 2014). Due to the LiDAR limitations referred to above, certain man-made features might only be detected through field prospection, inter alia by reference to the presence of characteristic ruderal vegetation-a valuable indicator of past human activities in abandoned areas (Latocha et al., 2019).
Last but not least, it is worth remembering that LiDAR itself does not ensure archaeological interpretations . Rather, it is the researcher's expertise and personal experience of the local landscape that underlies the process of appropriate reconstruction (Banaszek, 2014;Johnson & Ouimet, 2018). And DEM visualization, no matter how accurate, is still not a map. An interpretation key has to be provided, because 'interpreting means correlating trace with the event that produced it, supplementing the trace with a mental image of what is missing from it' (Mlekuž, 2013b). Only then, by contextualizing the traces of past societies imprinted in the relief, may LiDAR contribute vitally to the answering of anthropological questions regarding interactions between humans and the surrounding landscape (Johnson & Ouimet, 2014).

| CONCLUSIONS
We demonstrated that the mountain villages in Poland abandoned after World War II offer a unique opportunity for accurate reconstruction of past cultural landscapes dating back to the feudal period. The spatial pattern visible on mid-19th-century cadastral maps is still well conserved in the relief and detectable using LiDAR. We found that the landscape pattern imprinted in that relief was largely shaped by environmental conditions (lithology, soils and topography), the initial landownership division and land use. As permanent abandonment following post-WWII displacements of population was observed in mountains throughout Central and Eastern Europe, our case study can be considered representative of a regionally relevant phenomenon.
The approach taken allowed for the analysis of a range of LiDARderived landscape features, as further compared between study sites, and relate our results to those from other mountainous regions and landscapes. We provided new evidence regarding the visibility of past landscape features that can help build international comparative studies and syntheses.
Although laser scanning proves to be a powerful tool in landscape archaeology, it has its limitations like any science-based technology.
LiDAR's eye looks deeply and penetratingly, but the stand-alone tool does not understand what it perceives. Hence, accurate identification and dating of microtopographic man-made features is possible only when LiDAR data are integrated with data from other sources, including field survey. It is often believed that the main added value of laser scanning in archaeology is the possibility to reveal previously unknown objects, and accurately and precisely determine their position, size and shape, as well as evaluate their state of preservation or degradation. Notwithstanding the foregoing, we argue that the greatest advantage of LiDAR is its capacity to provide a landscape context to isolated traces of past human activities and thus allowing for the reconstruction of entire spatial patterns and interrelationships developed by past societies.