Introduction

Research on historical landscapes has received much attention in recent years due to the growing availability of historical maps in digital form and innovation in mapping methods (Coomans et al. 2019). One of the possibilities for gaining access to knowledge about historical landscapes is through the technique of airborne laser scanning (ALS) (Johnson and Ouimet 2018; Cody and Anderson 2021). As a result of using this technique, a point cloud is achieved, enabling a digital elevation model (DEM) representing land surface elevation to be produced. Different DEM visualisation techniques make detecting preserved landscape elements connected with past human activity possible, both for historical and prehistorical epochs. Hence, this technique could be helpful for historical geographers using the retrogressive method that aims to reconstruct past landscapes in a particular cross-section (Vervloet 1984).

Until now, mainly series of consecutive historical maps and aerial photographs have been used in research on landscape changes (Affek et al. 2020; Sobala 2022). The most striking advantage of cartographical materials is that they make it possible to perceive spatial characteristics and measure past landscape elements. As a result, it is possible to analyse changes in particular landscape elements, especially in land cover and land use. However, some authors have drawn attention to the need to verify the content of archival maps as they are often burdened with errors (Kienast 1993; Wolski 2012). Moreover, historical maps limit the period of research, usually maximally to the turn of the eighteenth and nineteenth centuries. Although it is sometimes possible to delve deeper into the past, the obtained results cannot be analysed quantitatively due to the low quality of the oldest maps. Furthermore, maps show the temporary state of the landscape indicated by the mapping time, and what happened between successive time periods is interpreted by the researcher and thus burdened with a certain amount of uncertainty. This could, for example, have resulted in a faulty assessment of the maximal human impact in the past in some areas (Sobala 2021).

There has been the possibility to verify the content of historical maps in new ways over the past 20 years, such as airborne laser scanning (ALS). As far as historical landscapes are concerned, this technique has mainly been applied in archaeological studies, but it has been also used by geographers (Affek et al. 2022). Similarly to historical maps, spatial measurement is possible, which is an advantage of great importance in past landscape studies. The application of ALS has revolutionised the practise of settlement and landscape archaeology. For example, it allowed the distribution of settlements in the Maya lowlands to be characterised for the first time (Ringle et al. 2021) and showed that the impact of human activities in the Polish part of the Białowieża Forest were substantially more significant than previously believed (Stereńczak et al. 2020). ALS data compared with aerial and satellite images, historical cartography and field surveys allowed unreported prehistoric and Roman features to be identified and detailed topographic information to be obtained about known sites in central northern Istria, Croatia (Bernardini and Vinci 2020). Recently, geographers have also started to use this technique to describe whole landscapes and their particular elements. Affek et al. (2022) demonstrated that using ALS in mountain villages in Poland that were abandoned after World War II offers a unique opportunity for accurate reconstruction of past cultural landscapes dating back to the feudal period. In turn, Duma et al. (2020) gave a comprehensive presentation of the stone walls in the Izera Mountains in Poland, including their morphometric characteristics, whereas Johnson and Ouimet (2016) characterised stone walls in New England in the USA. ALS data were also used to detect landscape elements connected with military activity during World War II (Jucha et al. 2021; Waga et al. 2022).

Nevertheless, some authors have noted the limitations of ALS data (Affek 2014). As varied techniques of DEM visualisation are applied to identify landscape elements connected with past human activity, they provide different results in the detection of landscape elements. Hence, there is a need to assess their usefulness. Some studies comparing the efficacy of visualisation techniques have been conducted for different land cover (Stular et al. 2012; Bennett et al. 2012; Kiarszys and Banaszek 2017). However, until now, there have been little studies on that issue in marginal mountain areas characterised by varied land cover and topography (Affek et al. 2022) and the usefulness of ALS data has not been assessed quantitatively in this type of area. Marginal mountain areas are those of little agricultural or developmental value where traditional agriculture was practised until recent decades (Nunes et al. 2011; Ahmadzai et al. 2022). They are characterised by poor soil quality, remote location, lack of hardened roads, difficulties in agricultural activity connected with slope inclination etc. Land abandonment and afforestation have helped preserve the traces of former human activity in these areas. As a result, the usefulness of techniques of DEM visualisation could be assessed. This paper aims to compare the usefulness of different DEM visualisation techniques for the purposes of detection of landscape elements connected with past human activity in marginal mountain areas in the Western Carpathians. Furthermore, some problems connected with ALS data application are pointed out.

Materials and methods

Study area

The study was conducted in the Western Carpathians (Fig. 1). There has been a constant increase in forest cover here since the turn of the nineteenth and twentieth centuries due to agricultural land abandonment. Afforestation following previously observed deforestation is typical of many mountain areas in developed countries and is known as the “forest transition” (Mather 1992). In the Western Carpathians, afforestation accelerated after World War II; however, in the higher parts of this mountain range, this process started at the end of the nineteenth century (Sobala et al. 2017).

Fig. 1
figure 1

Study area. A1—Radziechowska, A2—Jaworzyna, A3—Ostre, B1—Praszywka, C1—Prusów

Full size image

The origin of human activity in the higher parts of the Western Carpathians began in the fifteenth century. This activity resulted from the migration of the Vlach people in this area, whose activity was connected previously mainly with grazing. They founded new settlements at higher elevations and formed glades for sheep to graze by slashing and burning the forests (Jawor 2000). An increase in the population of mountain valleys led to a progressive increase in arable land range to higher parts of valleys, slopes and even mountain ridges. As a result, some mountain meadows and pastures were converted into arable lands and permanent or seasonal settlements (Broda 1956). Pastoral economy started to collapse in the nineteenth century, which was caused by many factors such as an intensification of the forest management and development of industry, enfranchisement of peasants, the elimination of servitudes and the need to pay for grazing in meadows and in court forests. In many cases, arable fields were also abandoned because the opportunities for employment in industry and services increased. Following this stage, the area of mountain pastures and glades decreased. This process has continued up to the present day (Kozak 2010; Sobala 2018). The agricultural land abandonment and the related afforestation phase have helped preserve the traces of former human activity, which could be destroyed if there is a new land use direction. These traces include anthropogenic terrain forms related to agricultural land use in the past (embankments of agricultural terraces, stone walls and mounds or balks) and preserved objects or the traces of them (ruins of farm and residential buildings, wells, roads) (Table 1, Fig. 2).

Table 1 Basic characteristics of terrain forms and objects related to past agricultural land use in the Western Carpathians
Full size table
Fig. 2
figure 2

Terrain forms and objects related to past agricultural land use in glades in the Western Carpathians. Photographs and DEM visualisation present: a embankments of agricultural terraces in Praszywka, b balk in Praszywka, c stone mound in Prusów, d stone wall in Ostre, e road in Radziechowska, f remains of a residential building in Prusów. DEM visualisation techniques applied: MHS: a–c, e–f, SVF: d (photograph by Michał Sobala)

Full size image

The detailed study was carried out in five test areas in the Silesian Beskids, Żywiec-Kysuce Beskids and Żywiec-Orawa Beskids (the Western Carpathians) (Fig. 1). To ensure a high level of representativeness for a large area, the chosen test areas are characterised by varied topographical conditions (elevation, slope, aspect) and varied current land cover (both contemporarily deforested and reforested as a result of forest succession) (Table 2). These test areas were also characterised by a different range of agricultural activity in the past. Grazing predominated in Radziechowska (A1); the upper parts of Jaworzyna (A2), Praszywka (B1) and Prusów (C1) were used as pastureland, whereas their lower parts as arable fields and meadows; land use in Ostre (A3) was mixed (Sobala 2021). Nowadays, only some parts of these test areas are utilised for extensive sheep grazing (Radziechowska) or seasonal settlement and moving (a few holiday huts in Jaworzyna and Prusów).

Table 2 Basic characteristic of test areas
Full size table

Materials

In this paper, the following materials were used: a digital elevation model (DEM) and a digital surface model (DSM), an orthophotomap, and a topographic map of Poland at a scale of 1:10,000.

The digital elevation and surface models used in this study are ready-made products based on ALS with a density of 4–8 points/m2. They were made in a projected coordinate system 1992 (EPSG: 2180) with a 1 × 1 m resolution and an average height error in the range of 0.2 m. The point cloud was acquired in 2020 within the national ISOK (Informatyczny System Osłony Kraju) project. The data were collected in a multi-return mode with a maximum of five return echoes registered for each emitted pulse and a field of view of less than ± 25°. Point cloud was classified according to the American Society for Photogrammetry and Remote Sensing standards into ground (class 2), low (3), medium (4) and high (5) vegetation, buildings (6), low points (7), model keypoints (8), water (9) and others. This classification was performed automatically using the progressive densification filtering algorithm and subsequently followed by a visual inspection and manual correction of misclassified points (Meng et al. 2010).

The orthophotomap with a pixel resolution of 0.25 × 0.25 m corresponds to a 1:10,000 scale map and is a raster cartometric result of the orthogonal processing of aerial photographs. It was made in 2020 in a projected coordinate system 1992 (EPSG: 2180). The topographic map of Poland was made in 1980 in a projected coordinate system 1965 (EPSG: 2175) by the civil service for economic purposes. The topographic map and orthophotomap were used for orientation during field mapping. All materials were received from the Head Office of Geodesy and Cartography (Główny Urząd Geodezji i Kartografii, GUGiK).

Detection of terrain forms and objects

The detection of anthropogenic terrain forms and objects was conducted in three stages. Firstly, reconnaissance was carried out to observe all the anthropogenic forms and objects occurring in the study area (Table 1, Fig. 2). Of all the observed anthropogenic terrain forms and objects, the ones connected with agricultural activity, such as embankments of agricultural terraces, balks, stone walls and mounds and roads, were chosen for further study. Secondly, DEM visualisation techniques were used to detect manually these forms and objects (Table 3; Fig. 3) (Kokalj and Hesse 2017). These visualisations are the most often used and implemented in open-source desktop applications. They were created using Relief Visualisation Toolbox, ver. 2.2.1. (Zakšek et al. 2011) and were interpreted then by one person in ArcGIS 10.8. The results of this interpretation were saved as shp files. As it was impossible to distinguish between some of them without field study, they were aggregated into four groups: (1) embankments of agricultural terraces including stone walls perpendicular to a slope, (2) balks including stone walls parallel to a slope, (3) roads including stone walls perpendicular to them and (4) stone mounds. The density of these forms detected in particular test areas is presented in Table 4. As different visualisation techniques provide varied results, the range of density was presented.

Table 3 DEM visualisation techniques applied in the study (Kokalj and Hesse 2017)
Full size table
Fig. 3
figure 3

Terrain forms visible on DEM visualisation techniques applied in the study: a embankments of agricultural terraces, balks and stone mounds in Praszywka, b part of road in Radziechowska

Full size image
Table 4 The range of density of analysed terrain forms calculated based on six DEM visualisation techniques [m/ha]
Full size table

The last stage of the study was the field mapping of these forms and verification of the results of the DEM interpretation. It must be emphasised that using DEM visualisation techniques in the second stage was helpful for field mapping. However, the aim of field mapping was also to reject all the terrain forms and object that were falsely detected based on DEM visualisation techniques. During field mapping a GPS receiver GPSMAP®66i was used. Stone mounds as point objects were detected using point marking, whereas the LiveTrack tool was used to detect linear forms. All data were saved in gpx files that were subsequently converted into shp files using the GPX to Features tool in ArcGIS 10.8. During the terrain inventory, the orthophotomap and the topographic map of Poland were used.

After detecting terrain forms and objects using DEM visualisation techniques and field mapping, the length of linear forms was measured, and the number of point forms was calculated for each detection method. These parameters were calculated separately for each test area, taking into account the character of vegetation that could influence the results obtained: low vegetation (< 0.5 m) and high vegetation (≥ 0.5 m). The height of vegetation was given from the normalised digital surface model (nDSM).

Unfortunately, it was impossible to measure all linear forms during field mapping due to the unavailability of some areas because of dense vegetation. This problem mainly concerned embankments of agricultural terraces and balks. In the case of roads, almost all their stretches were measured because of their accessibility. As far as embankments of agricultural terraces and balks are concerned, the field mapping aimed to verify the detection results based on DEM visualisation techniques.

Assessment of visualisation techniques

To compare the usefulness of the different DEM visualisation techniques in the detection of linear forms, the statistical method was used. The aim of the analysis was to examine the effects of terrain form type, vegetation type, and DEM visualisation technique on length change and significance among categories within predictors. The main purpose was to determine which type of DEM visualisation technique produced significantly better results in terms of length for each type of terrain form.

A linear multilevel model was fitted (estimated with REML and nloptwrap optimizer) to predict length with terrain form, vegetation and DEM visualisation technique. The model included test area as random effect. In this model, the length variable, Yij (x ∈ ℤ+) was a function of three nominal covariates: terrain form variable [“roads”, “agricultural terraces”, “balks”] (Table 1), the DEM visualisation technique variable [“MHS”, “PCA”, “SVF”, “LRM”, “OPN”, “OPP”] (Table 3) and the dichotomous vegetation variable [“Low”, “High”] which vary across test areas i [“A1”, “A2”, “A3”, “B1”, “C1”] (Fig. 1, Table 2). Nominal variables were specified by dummy coding.

The model was specified either by formulations at both levels:

  • Level one: \({Y}_{ij}= {a}_{i}+b\bullet {\mathrm{terrain form}}_{i}+c \bullet {\mathrm{technique}}_{ij}+d\bullet {\mathrm{vegetation} }_{i}+e\bullet {\mathrm{terrain form}}_{i}\times \bullet {\mathrm{technique}}_{ij}+{\epsilon }_{ij}, \mathrm{where} {\epsilon }_{ij}\sim N\left(0, {\sigma }^{2}\right)\)

  • Level two: \({a}_{i}= {\alpha }_{0}+{u}_{i}, \mathrm{where} {u}_{i} \sim N\left(0, {\sigma }_{u}^{2}\right)\)

  • Or as a composite model: \({Y}_{ij}= {\alpha }_{0}+b\bullet \mathrm{terrain form}+c \bullet {\mathrm{technique}}_{ij}+d\bullet {\mathrm{vegetation} }_{i}+e\bullet {\mathrm{terrain form}}_{\mathrm{i}}\times \bullet {\mathrm{techniqu}e}_{ij}+{u}_{i}+{\epsilon }_{ij}\) (mod.1).

The ai was the intercept for roads measured by MHS technique with low vegetation; b was the difference at baseline for those terrain forms with experimental condition [“agricultural terraces”, “balks”] vegetation relative to those in the control condition (technique =  = “roads”); c was the difference at baseline for those length with experimental technique [“PCA”, “SVF”, “LRM”, “OPN”, “OPP”] relative to those in the control condition (technique =  = “MHS”); d was the difference at baseline for those length with high vegetation relative to those in the control condition (vegetation =  = “Low”). d was the terrain forms by technique interaction, which showes the difference in change for those in the experimental condition, relative to those in the control condition. The significance of the differences in this measure allowed to test the research hypothesis.

Parameter ai was allowed to vary between test areas. The level-2 equation shows that ai was decomposed into a large mean α0 and location-specific variances around this large mean ui. The \({u}_{i} \sim N(0, {\sigma }_{u}^{2})\) line illustrated that the conventional assumption that these locations-specific deviations were normally distributed was made, with a mean of zero and a standard deviation σu of level 2. Thus, between-location variability was captured in σu and within-location variability in σϵ.

The significance level of the statistical tests was set at α = 0.05.

Analyses were conducted using the R Statistical language (version 4.1.1; R Core Team 2021) on Windows 10 × 64 bit (build 19,044), using the packages lme4 (version 1.1.27.1; Bates et al. 2015), Matrix (version 1.3.4; Bates and Maechler 2021), emmeans (version 1.7.2; Lenth 2022), ggeffects (version 1.1.1; Lüdecke 2018), sjPlot (version 2.8.10; Lüdecke 2021), report (version 0.5.1.3; Makowski et al. 2021).

Point terrain forms, such as stone mounds, were not statistically analysed as their number detected based on ALS data was very low compared to the number of forms detected during field mapping. These terrain forms are of different sizes. The largest forms are up to 3 m high; nevertheless, many stone mounds are much smaller (20–40 cm), and their shape is flattened. In such cases, these forms are not visible in the DEM visualisation or are difficult to distinguish unequivocally.

Results

Comparison of visualisation techniques for the detection of linear forms

Using the composite model specification the mod.1 was fit with the results shown in Table 5. The model’s total explanatory power was substantial (R2conditional = 0.83) and the part related to the fixed effects alone (R2marginal) was of 0.29. The model’s intercept, corresponding to terrain form = roads, vegetation = low and technique = MHS, was at 3065.94 (95% CI [1545.71, 4586.17], t(135) = 3.99, p < 0.001).

Table 5 Results of fitting mod.1, Nobs = 156
Full size table

No significant differences in length of terrain forms were found between the DEM visualisation techniques for any of the terrain forms, leading to the conclusion that among the techniques studied, there is none that provides significantly better length results compared to the others for the terrain forms studied.

Predicted values for length based on terrain forms, DEM visualisation techniques and vegetation are shown in Fig. 4. Predictions of average length values (dots) within each terrain form were very close for all DEM visualisation techniques, and the overlap of confidence intervals (vertical lines) graphically illustrates the lack of significant differences.

Fig. 4
figure 4

Prediction of length (m) by fitted mod.1 with terrain forms, techniques terms

Full size image

In the case of roads, there was the possibility of measuring their length during field mapping. Hence, the comparison of results obtained using DEM visualisation techniques with the length measured during direct field observation was possible. The length of the roads measured using the interpretation of DEM visualisations ranges from 65 to 82% of the length of the roads measured during field mapping. On average, the best result was obtained for OPN.

Difficulties in the detection of forms and objects

There are some outliers in the results obtained for different test areas that could reflect varied natural conditions which influence the usefulness of DEM visualisation techniques in mountain areas. These natural conditions have limited possibility to use different visualisation techniques and GPS detection during field mapping.

For example, in the A1 test area, the usefulness of road detection during filed mapping in areas covered with high vegetation was lower than in other test areas. Due to vegetation, roads were inaccessible and their terrain measurement impossible. Simultaneously, these roads were visible on ALS data (Fig. 5a). In turn, in the B1 test area, the usefulness of roads detection during field mapping was lower in the area covered with low vegetation compared with other test areas. Dense low grass vegetation made roads detection difficult (Fig. 5b).

Fig. 5
figure 5

Difficulties in road detection during field mapping caused by different types of vegetation: a high vegetation in Radziechowska makes the area inaccessible, b low vegetation in Praszywka obscures roads. Explanation: 1—roads detected during field mapping, 2—roads visible on ALS data that were not detected during field mapping. DEM visualisation techniques used: OPN for a and OPP for b (photograph by Michał Sobala)

Full size image

As mentioned earlier, in the case of roads, their measurement during field mapping was possible in most cases. In turn, visualisation techniques gave the only possibility to measure embankments of agricultural terraces and balks. Admittedly, part of these forms were located in areas covered with low vegetation (Fig. 6a) or in areas covered with forests with sparse undergrowth (Fig. 6b), and their measurement was not difficult. Nevertheless, some of the areas covered with forest hindered terrain form measurement (Fig. 6c–f). In such cases, it was only possible to confirm their occurrence during field mapping.

Fig. 6
figure 6

Location of selected linear terrain forms according to land cover: a embankments of agricultural terraces in areas covered with grasses did not cause a problem during field mapping (Prusów), b stone walls edging past arable fields in a forest where measurement was possible (Ostre), c stone wall emerging form forest that could be measured only partially (Prusów), d stone wall that could be detected, but its measurement was impossible due to tree cover (Ostre), e spruce bosk that made measurement of linear forms during field mapping impossible (Praszywka), f dense forest vegetation with downed trunk making measurement of linear forms during field mapping impossible (Prusów). (photograph by Michał Sobala)

Full size image

Contrary to linear forms, stone mounds of various sizes (height ranges from several dozen centimetres to above two metres) were difficult to identify unambiguously based on the DEM visualisation techniques, especially within forest-covered areas. The field mapping gave a much better result, also allowing the identification of small stone mounds. In addition, field mapping made distinguishing stone mounds from other point forms, such as piles of brushwood gathered in a forest, possible.

Discussion

For a long time, research on past landscapes has focused on the remains of cultural heritage detected based on historical maps or detailed field mapping in small areas. The situation has undergone significant changes, with access to new remote sensing techniques, such as airborne laser scanning (Affek 2016). This and growing access to ALS data in many countries (Holata and Plzák 2018) have resulted in increasing interest in such research (Johnson and Ouimet 2018). However, in many cases, ready-made data were developed for other aims than detecting remains of cultural heritage. Hence, the identification of small linear and point terrain forms based on ready-made ALS data could be limited because of the inability to control or verify the correctness of DEM modelling (Kiarszys and Szalast 2014). As a result, there is a need to assess the usefulness of ready-made ALS data.

This study showed that, as a rule, the usefulness of particular DEM visualisation techniques is similar. This relates to the research results of Kiarszys and Banaszek (2017), which show the lack of possibility to indicate the best DEM visualisation technique. It is difficult to point unambiguously to the most useful among all the analysed techniques. This confirms the statement that a variety of terrain forms and objects and their locations in varied geomorphological conditions require the use of at least a few visualisation techniques (Kokalj and Hesse 2017).

Nevertheless, among all the analysed DEM visualisation techniques, OPN gave the best results in road detection. These results confirm that this technique is appropriate for detecting concave terrain forms and objects such as roads (Banaszek 2015). In turn, the best result for embankments of agricultural terraces and balks was obtained using OPP. This confirms that this technique is appropriate for the detection of convex terrain forms and objects (Banaszek 2015). However, it must be emphasised that the relatively high results of the detection of terrain forms and objects using OPP and OPN could be caused by the fact that these techniques were applied last. As was pointed out by Kiarszys and Banaszek (2017), the interpretation of different DEM visualisation techniques, to some extent, could influence each other, even though they are conducted separately. This is because an interpretation of the subsequent DEM visualisation techniques is preceded by obtaining information on the location of the interpreted objects based on the previous techniques.

The most crucial disruptive factor in the detection of terrain forms during field mapping was vegetation cover. Measurement of embankments of agricultural terraces and balks during field mapping was impossible in many cases because of dense vegetation. This did not concern roads as many of them are still extensively used by tourists, foresters and residents. Vegetation influences also a usefulness of DEM visualisation techniques. Firstly, it increases the scattering of light during flight and decreases the number of reflections from the ground (Opitz 2018). Secondly, there could be a problem during point classification in distinguishing reflections from low vegetation and ground, especially if the topography is varied. This mainly concerns tiny point terrain forms (Guan et al. 2014) and could be the reason why the percentage of stone mounds detected using the DEM visualisation techniques in this study was low. As was pointed out by Crow (2009), ALS based identification is more problematic for point features than linear features. Point features can easily be obscured by dense vegetation or omitted where point cloud density is too low. Furthermore, other studies have shown that it is possible to mistake them for other features such as piles of decomposing branches (Doneus et al. 2008; Affek 2014; Schindling and Gibbes 2014), tree boles, stumps and wood dump sites (Horňák and Zachar 2017).

As a result, this study has confirmed that using ALS data and field mapping should be complementary. First of all, there is a need to verify anthropogenic terrain forms during field mapping (Affek 2016). This is because not all existing terrain forms can be clearly identified based on DEM visualisation techniques. However, without identifying these forms based on DEM, it would be difficult to recognise them during field mapping. Due to secondary forest succession in abandoned glades, some forms are masked by vegetation, and reaching some parts of the glades was difficult, sometimes even impossible. In those cases, these terrain forms could only be measured based on ALS data.

The study was conducted in five varied test areas that, on the one hand, enabled different topography and land cover conditions that are typical of mountain areas to be taken into account. On the other hand, it might have influenced the diversity of results obtained in particular test areas. This is because DEM visualisation techniques depend not only on the characteristics of the detected features (their size, shape, convexity, concavity) but also on the topography (Kokalj and Hesse 2017). Different techniques can preferentially emphasise small-scale depressions or elevations, low relief features on horizontal or sloping planes, and structures on slopes. Some work better on almost flat terrain, or topography with gentle slopes (LRM), while others work better on rugged or mixed terrain (SVF) (Kokalj et al. 2018). Hence, more studies on this issue should be conducted in mountain areas. All in all, using ALS data do not give the full confidence about occurrence of terrain forms and objects but rather about their visibility on DEM visualisations (Affek et al. 2022).

While the use of ALS is common, the results obtained show that it is necessary to be aware of uncertainty and errors when using ALS data. As mentioned earlier, none of the DEM visualisation techniques provides complete information on traces of past human activity. However, there are a lot of prospects for developing the usefulness of ALS data application in research, such as changes in flight parameters (Affek 2014), combination and overlapping of different visualisation techniques (Novak and Oštir 2021), their matching to local conditions (Kokalj et al. 2018), the application of point cloud to control the obtained DEM and own point classification (Opitz 2018) and the use of other complementary materials such as archival maps or aerial photographs (Affek et al. 2022).

Conclusions

The conducted study shows that there is no possibility to point out the most useful DEM visualisation techniques for detecting anthropogenic terrain forms and objects connected with past human activity in marginal mountain areas. The average usefulness of applying particular DEM visualisation techniques does not differ significantly; however, this effectiveness varies slightly in different topography and land cover conditions. The results obtained confirmed that OPN is useful for detecting concave terrain forms and objects such as roads. In contrast, OPP is appropriate for detecting convex terrain forms and objects such as embankments of agricultural terraces and balks. More studies are needed to assess the influence of topography on the results obtained using different DEM visualisation techniques.

Furthermore, this study proved that in marginal mountain areas characterised by secondary forest succession, only DEM visualisation techniques provide the possibility to measure the length of detected embankments of agricultural terraces and balks on abandoned land. Measurement during field mapping is impossible due to dense vegetation cover. This conclusion concerns roads to a lesser extent because they are mostly still used, and thus they are not covered by dense vegetation in many cases. On the contrary, the study proved the minor usefulness of DEM visualisation techniques in the detection of point terrain forms such as stone mounds. It was impossible to detect these forms using the DEM visualisation techniques in most cases. This is connected with the varied size of these forms; only the biggest forms were visible in DEM.

In this study, the research sample was small, which was connected with the application of the time-consuming terrain verification of all the results obtained during DEM analysis. However, this emphasises the necessity to look for less time-consuming methods for collecting data on past landscape elements, especially in mountain areas characterised by harsh topography conditions. Furthermore, this highlights the necessity to assess the usefulness of applying ALS data in varied topographical and land cover conditions and the need to be aware of the limitation of using ready-made ALS data. As the usefulness of ALS data in detecting terrain forms and objects depends on applied data processing techniques, customising the entire process of generating DEM visualisations will allow for more accurately detection.