In search of ancient pre-Roman imperial roads: state of research and some methodological recommendations

In recent decades, there has been a considerable growth of scientific interest in the question of travel and mobility in the ancient Mediterranean and Near East. As far as the southern Levant is concerned, this interest has yielded much research about travel and mobility in Late Antique Palestine (including the Roman imperial road network), leaving the question of pre-Roman travel and mobility a much-understudied field of study. However, recent archaeological discoveries have revolutionized our state of knowledge about material remains of pre-Roman roads, which were once believed to not have been preserved. Namely, remains of several pre-Roman roads have been discovered on the ground in modern Jordan and Israel: the Aroer Ascent, “Glueck’s Road,” Naqeb Dahal, the Wadi Zarqa-Main road, the Callirrhoe–Machaerus road, and the Masada–Hebron road. In this context, the aim of this paper was, first, to study the spatial and archaeological features of the archaeologically attested ancient pre-Roman roads, and second, to evaluate all of the available types of remote data as tools serving in the detection of artifacts (especially archival cartographic, aerial, and satellite data; archaeological data; and least-cost simulations of Geographic Information Systems [GIS]). As a result, several conclusions are suggested. First, old topographic maps are a very promising source of data for potential new discoveries since they contain both explicit and implicit hints at the existence of ancient roads. Second, information about archaeological data and the Roman imperial road network may also offer important reference points if the quality of databases is improved in the future. Third, the use of GIS simulations based on 30 m digital elevation models has very substantial limitations. In particular, the use of least cost paths and corridors for simulating short-distance travel has turned out to be fruitless. At the same time, the use of least cost corridors for predicting long-distance travel offers some promising results.


Introduction
In recent decades, there has been a considerable growth of scientific interest in the question of travel and mobility in the ancient Mediterranean and the Near East. As far as the southern Levant is concerned, this interest has yielded one fundamental publication about the Iron Age roads in ancient Israel (Dorsey 1991) and has also generated significant scientific output on mobility in Late Antique Palestine (e.g. Elsner and Rubiés 1999;Elsner and Rutherford 2008;Hezser 2011;Ellis and Kidner 2011;Yoo et al. 2018). In contrast, it should be stressed that except for Jewish pilgrimage (Safrai 1981;Dyma 2010), the question of travel and mobility in its multifaceted complexity in Palestine in the period between the Iron Age and Late Antiquity remains a much-understudied field of science.
What is more, it was long and commonly held that no extant archaeological evidence existed for interurban ancient roads before Roman Imperial paved roads in the ancient Near East (Beitzel 1991) and in ancient Palestine in particular (Dorsey 1991). However, archaeological discoveries in recent decades have revolutionized our state of knowledge in this regard. First, as far as the Near East in general is concerned, the past two decades have given rise to the discovery of many of what have become known as hollow ways in the Middle East, especially Syria and Iraq (Ur 2003;Wilkinson et al. 2010;de Gruchy and Cunliffe 2020). Hollow ways are broad and shallow linear depressions in the landscape that are thought to be formed by the continuous passage of human and animal traffic (Ur 2003). Hollow ways are usually first detectable on satellite imagery as unusual geometric concentrations of vegetation or moisture-unusual in the sense that they are not justified by the current land use. Second, remains of several pre-Roman roads were discovered on the ground in modern Jordan and Israel: the Aroer Ascent, "Glueck's Road," Naqeb Dahal, the Wadi Zarqa-Main road, the Callirrhoe-Machaerus road, and the Masada-Hebron road (Strobel 1997;Kloner and Ben David 2003;Ben David 2009Ji 2019;Davidovich et al. 2022).
One of the ancient roads that has recently been identified is located in Wadi Mujib in modern Jordan (Kloner and Ben David 2003;Ben David 2009). Wadi Mujib is a deep canyon (ranging from 250 to 800 m in depth) crossing the center of the Moab Plateau-it runs east from the northern end of the Dead Sea for about 40 km. The canyon was called the Arnon River in biblical and Roman times and became known as Wadi Mujib starting in the early Muslim period. Given the canyon's difficult terrain, it could only be crossed on foot along the 3-4 km of its length in the area marked by the existence of three Iron Age sites: Dhiban, Aroer, and Lehun. While the Dhiban ascent is the location of both Roman Imperial (Via Nova Traina) and modern roads, which likely covered the earlier ancient roads, the Aroer Ascent features the remains of a wide road, constructed in some sections with curb stones and retaining walls. The width of this road is sometimes between 4 and 6 m, and at times, the road turns into a regular camel naqab. Given the dating of Tell Aroer to the Iron Age, the Iron Age period has also been suggested for the dating of this road.
The case of an ancient road in Wadi an-Nukhayla is somewhat irregular (Kloner and Ben David 2003;Ben David 2009). Nelson Glueck, one of the pioneers of biblical archaeology in modern Jordan, published in his books an aerial photograph taken in 1937 by the British Royal Air Force showing impressive remains of what was clearly an ancient road. The photograph was accompanied by the caption "Roman road on north slope of Wadi Mujib" (Glueck 1939(Glueck , 1940 or "Roman road ascending Wadi Mujib" (Glueck 1965). No precise details for its location were given (Glueck 1965), and this road, although never actually ground-truthed, was interpreted as Roman Imperial and thought to have been located in Wadi Mujib, crossing it from north to south in the center of the Moab Plateau (Kennedy 2000). Remarkably, this road was accidentally localized on the ground in 2000 in Wadi an-Nukhayla (the southern branch of Wadi Mujib), some 12 km south-southeast of the expected location in Wadi Mujib in the center of the Moab Plateau (Kloner and Ben David 2003). The road is reported to have geomorphological features different from Roman Imperial roads, especially the lack of pavement (built on packed dirt), the arrangement of curb stones in rows along its way that was not built into the wall, and an irregular width of often less than 4 m (Ben David 2009). What is more, on top of the ascent, there are remains of an elliptical structure and an adjacent tower. The site is known as Rujm al-ʿAbid among the local Bedouin. The structure is dated to the Iron Age on the basis of collected pottery and architectural features and is consequently interpreted as the remains of an Iron Age fort (Parker 1987). In a broader topographical context, it may be speculated that this road served the communication between Rujm al-ʿAbid and the site of Balu in the southwest of the Kerak Plateau.
A path called Naqab Dahal is attested in literature as a local road from at least the times of Lawrence of Arabia (who used it in his raid from Beer Sheba to Tafilah), but it was explicitly identified as an ancient road only recently due to both the geomorphological features and the existence of an Iron Age fortification structure along the path (Ben David 2009). Given its location, Naqab Dahal apparently served communication between ancient Bozrah and Beer Sheba via Ein Hatzeva in the Arabah Valley.
An important ancient road connecting two well-known Herodian sites, Callirrhoe and Machaerus, was known as early as in the nineteenth century CE, but only recent studies have brought more clarity on its course and dating (Ben David 2015). Given the fact that Machaerus was destroyed in 72 CE and abandoned afterwards, this road must be dated to the Early Roman (Herodian) period at the latest. Its width exceeds 2 m along its entire length, and the road occasionally features curb stones and retaining walls.
Another important road (with a length of about 12 km) is called the Wadi Zarqa-Main road (Strobel 1997;Ji 2019). It connects the ancient site of Ataruz in the Moab Plateau with Jericho in the Jordan Valley via the important site of Boz al-Mushelle. The course of this road is marked by several archaeological sites, especially watchtowers. The ancient road is partly detectable on the ground, including the remnants of retaining walls and stone steps, and partly via the course of the modern dirt road. Based on the pottery finds and architectural remains of the sites related to the road, it has been proposed that the road was constructed in the ninth and eighth centuries BCE and was again extensively used from the second century BCE to the first century CE.
In turn, another ancient road can be found in the Nahal Tzalim in modern Israel (Porat et al. 2016). The road was suggested to have served travel between Hebron and Masada in the Early Roman period. What is more, given the presence of two Roman military structures along its course at Mazad Bedar and Mazad Tsipirah, it was clearly adopted and used by the Roman legions, especially between 66 and 73 CE. The road is of various widths (from 1 to 6 m) and occasionally features later Roman paving and retaining walls.
Thus, the overview of the current state of research makes clear that in contrast to "the received wisdom," it is possible to discover interurban pre-Roman roads, and one must not simply assume that they cannot be discovered. At the same time, one may have the impression that most earlier discoveries in Israel and Jordan took place either by chance or at least did not result from a well-known systematic strategy that other scientists could easily follow. Furthermore, the role of remote sensing in the detection of artifacts was limited or nonexistent (not to mention the use of geophysical methods employed with success for similar research in the territories of ancient Greece and Italy; Tsokas et al. 2009;Mozzi et al. 2015;Rizzo et al. 2018). In contrast, most of these discoveries in Israel and Jordan took place during fieldwork, unlike the discovery of hollow ways in Syria and Iraq. In this light, the paper sets out to suggest, for the first time ever, a working methodology of remote sensing research that could be effective in the future search for ancient pre-Roman interurban roads in the southern Levant. In particular, two specific goals are set for this study: first, to study the spatial and archaeological features of the ancient pre-Roman Imperial roads that have already been discovered, and second, to evaluate all of the available types of remote data as tools serving in the detection of artifacts. As a result, ancient pre-Roman Imperial roads in the southern Levant will be better understood as archaeological and geographical phenomena, and an evaluation of various types of data and remote sensing methods applicable to future discoveries will also be suggested. As far as the geographical framework is concerned, the geographical area of interest matches the area sometimes labeled as the land of the Bible, including the modern territories of Israel, the Palestinian territories, and Jordan in the southern Levant, but without the area of the southern Negev (which may be seen as a highly distinctive subfield of science connected with the Nabateans).

Study area
The study area is located in modern Israel, the Palestinian territories, and Jordan in the Dead Sea catchment area (Fig. 1). The central point of this area is determined by the coordinates of 31°15′ north latitude and 35°30′ east longitude. The detailed study sites are situated around the Dead Sea at distances ranging from about 4.5 km (no. 2) to about 30 km (no. 5) in a straight line from the modern shoreline of the reservoir. On the eastern side of the Dead Sea are two large river valleys (about 90 km each): the Mujib Valley and the Al-Hasa Valley. The average elevation of the entire area is 370 m a.s.l.; altitude differences are more than 1200 m, with the highest point being Jabal al Ataitah (1640 m a.s.l.) and the lowest formed by the surface of the Dead Sea (− 430 m a.s.l.). The surroundings of the Dead Sea have a hot desert climate, with less than 50 mm mean annual rainfall. The average summer temperature is between 32 and 39 °C, and average winter temperatures range between 20 and 23 °C (Israel Meteorological Service, 2022).
From a historical point of view, the political affiliation of the area in question has changed over the course of history (Grabbe 2004(Grabbe , 2008(Grabbe , 2020. In the Iron Age (roughly 1200 BCE-500 BCE), the area west of the Dead Sea was under the control of the Judahite kingdom, with its capital in Jerusalem (including both the northern Negev in Beer Sheba and the Arad Valley and Judean central hills). In turn, the area east of the Dead Sea was home to the ancient cultures of Ammon (north of the modern Wadi Mujib/ ancient Arnon), Moab (between the modern Wadi Mujib/ ancient Arnon in the north and modern Wadi al-Hasa/ ancient Zared in the south), and Edom (south of modern Wadi al-Hasa/ancient Zared in the south). In turn, after 332 BCE, the situation in the region changed considerably. The area west of the Dead Sea first belonged to the ancient Idumeans (ca. 332-108/107 BCE) and next to the Judean kingdom after the conquest of Idumea by the Hasmonean dynasty in ca. 108/107 BCE. Likewise, the area of ancient Ammon was also conquered by the Judean kingdom, and this province was called Peraea in the Early Roman period (63 BCE-70 CE), while the territory south of the modern Wadi Mujib/ancient Arnon belonged to the Nabateans.
The importance of the road network in the study area should be emphasized. First, the road network in the study area was important both on an international and domestic scale. Namely, two important international arteries were located in the modern territories of Israel, the Palestinian Authorities, and Jordan (Isaac and Roll 1982): the King's Highway and the Way of the Sea. The former led east of the Jordan River and connected Damascus with the Red Sea; it was important for south Arabian commerce in particular. The latter ran along the Mediterranean coast and connected Syria with Egypt. Many other local roads that both connected the international arteries and served local needs also existed in the study area. In fact, it should also be stressed that the role of the road network in ancient societies cannot be underestimated (Hezser 2011). In social terms, the road network was of the utmost importance for enabling communication and the distribution of information and knowledge in oral societies. In economic terms, the existence of a well-developed road infrastructure was a condition for local farmers to be able to sell their surplus and for the entire region to profit from international trade. Regarding culture, it is widely accepted that economic contacts were closely tied with intercultural exchanges. As for religion, it may have had both an encouraging and restricting effect on people's mobility.

Source data
There are various hints in the literature indicating how one should speculate about the course of ancient roads (Aharoni 1979;Dorsey 1991;Faust and Erlich 2011), but this general knowledge has never been turned into a large-scale, systematic research attempt. Based on the literature and our own experience, a wide variety of data may tentatively be suggested as potentially useful for identifying ancient roads, especially archival cartographic sources, archival aerial imagery, archival satellite imagery, high-resolution satellite imagery, databases of archaeological sites, datasets of Roman Imperial roads, and various GIS simulations based on a digital elevation model (DEM).
Firstly, the data sources used in this study can be divided into analog (nineteenth-and twentieth-century topographic maps) and digital (aerial and satellite images and digital elevation models). The archival topographic maps used for the analysis came mainly from the collections of the Palestine Exploration Fund (2022) and the German Society for the Exploration of Palestine (2022). Aerial and satellite images were obtained from publicly available websites such as the APAAME project (2022), MEGA-Jordan database (2022), Pleiades project (2022), and Bing Maps Aerial (Zoom Earth, 2022). As for the altitude data, two freely available global DEMs were used: SRTM (Shuttle Radar Topography Mission) in version 3 (Farr et al. 2000) and ASTER (Advanced Spaceborne Thermal Emission and Reflection Radiometer). However, in order to maintain thematic consistency, it was decided to discuss individual data sources by grouping them into three main sets: (a) topographic maps, (b) aerial imagery and satellite images, and (c) digital elevation models.

Topographic maps
Topographic maps in the form of 7 separate sources were collected for the analysis ( Table 1). Some of them come from the nineteenth century, and although they are imperfect in terms of cartometry, they are particularly valuable because of their content. Old maps of the Dead Sea basin are extremely useful because they show the historical settlement system and were prepared by researchers focused on the monuments of the past. Newer topographic maps from the twentieth century, despite representing the modern settlement network, are also a very valuable source because the errors concerning the location of elements on the maps are relatively small. Topographic maps were collected from opensource archives such as the Palestine Exploration Fund collections (2022) and German Society for the Exploration of Palestine documents (2022). The associations responsible for the creation of these sources, which were founded in the nineteenth century, are currently archiving and disseminating them.
The oldest topographic source is the Map of Western Palestine created by the Palestine Exploration Fund (PEF) in 1872-1880 (Conder and Kitchener 1880). The survey of Palestine was published as a 1:63,360 map covering a range of 26 sheets. This work is especially valuable for researchers because it identifies a great number of places from biblical times. It is the most detailed and accurate map drawn of this region in the nineteenth century and, as Masalha (2019) states, the survey organized by the PEF was the peak of cartographic achievement in the Near East before the twentieth-century advent of topographic maps based on cadastral sheets. The great advantage of this source is its completeness in terms of coverage and consistency of execution between sheets.
The Survey of Eastern Palestine prepared in 1881 (Conder 1881) represents the continuation of the PEF's cartographic mission. The Eastern Palestine survey led by Captain Conder collected information from the area east of the Jordan River between Wadi Zarqa Ma'in in the south and Wadi el Hammam in the north. The map consists of only one sheet drawn at a scale of 1:63,360. In terms of map content, it is similar to the Western Palestine survey. Unfortunately, the mission could not be continued in the larger area of modern Jordan for political reasons (Conder 1889).
A rich source in terms of cartographically recorded archaeological objects east of the Jordan River is the Karte des Ostjordanlandes drawn up by Gottlieb Schumacher (1890), which is similar to the PEF's maps in terms of scale (1:63,360). The map consists of 8 sheets and covers the area from the Banias River in the north to the Zarqa River in the south. It is worth emphasizing that the sheets of this map from the late nineteenth century contain many sections of roads identified as Roman (ger. Römerstraße) or old (ger. Alte Straße). The archaeological interest of the mapping engineer is also evident in the numerous details identifying ancient ruins (Schumacher et al. 1889).
The first map from the beginning of the twentieth century, Arabia Petraea by Alois Musil, is a completely different elaboration. Completed in 1906, the map was drawn at a scale of 1:300,000 as a comprehensive cartographic approach to the region. In terms of the Dead Sea catchment, it covers the area on the east bank from Wadi al Hisban in the north as far as ancient Buseira in the south. The map was the first scientific survey of Nabataean remains (Musil 1907).
Another important cartographic source is the Map of Palestine (1918) created by the German Vermessungsabteilung for the purpose of military operations. The map, comprising 40 sheets at a scale of 1:50,000, was developed on the basis of the PEF maps and interpretations of aerial photographs in 1918 (Demhardt 2021). The map contains valuable information about roads, topography, and the crossing of large rivers that was collected for strategic reasons. Due to the  (1946), drawn by the no. 1 Base Survey Drawing and Photo Process Office, is a slightly later work. It is similar to the previous German maps in terms of thematic layers and scale of execution. The cartographic source is based on the Palestinian grid coordinate system (EPSG: 28,191), which was planned in 1922 (Gavish 2005). The map sheets are extremely rich in terms of content and contain information that is simultaneously important from both a military (bridges, river crossings, and landing sites) and archaeological point of view (ruins, remains of Roman roads, and the names of tells). Work on the preparation of the map sheets lasted until 1946.
The most modern source in cartometric terms is the Map of Israel (1956), developed at a scale of 1:100,000 in 1942-1956, which bears the hallmarks of contemporary topographic studies (Levin et al. 2010). The coordinate system of the Palestinian grid was also used in its production, which was organized by the Survey of Israel. The content of the map includes roads classified into 4 categories and, despite the large scale, local sections are also included. Apart from the contour lines, the sheets have been enriched with a hydrographic network that even considers natural sources of fresh water. The rich content of the map is the result of the generalization of more detailed 1:20,000 cadastral maps, which were worked on until the 1960s (Gavish 2005).

Aerial and satellite imagery
Archival aerial imagery is another source of spatial data that is invaluable in non-invasive, large-scale archaeological prospection (Hanson and Oltean 2013). The first aerial photos from the research area were taken by British and German aircrafts in the beginning of the twentieth century. However, there are only a few of these photos, and their quality also leaves much to be desired. Nonetheless, one of the discovered ancient roads was photographed during an archaeological reconnaissance conducted by the British military in 1936 (Glueck 1937). Many good-quality photos of the examined objects were obtained from the Aerial Photographic Archive for Archaeology in the Middle East project (APAAME, 2022), which effectively collects archival data and creates new collections of aerial photos. Similar archives of historical aerial imagery relevant to the territory of Israel are the Map Library and Aerial Photograph Archive of the Department of Geography of the Hebrew University (2022) and the Digital Media Center (2022) of the University of Haifa.
Another very specific kind of data is archival satellite imagery (Parcak 2009). Its great value lies in the fact that these images show large areas of the landscape from before the advent of urbanization and industrialization in the Middle East. Of special importance are satellite images that were collected by US intelligence aerial and satellite missions during the Cold War, especially CORONA satellite imagery from 1959-1972(Ur 2003. Their use requires a trained eye but has proven successful in satellite archaeology in the Middle East. A dataset of CORONA images was obtained from the repository of the Corona Atlas & Referencing System (2022) provided by the Center for Advanced Spatial Technologies.Archival images are available for the surveyed area in the form of a mosaic created on August 13, 1968, andJune 8, 1970. All of the images in the mosaic have already been georeferenced, and only the contrast (gamma value) needed to be manipulated to better visualize linear structures.
High-resolution satellite images are undoubtedly the most valuable source of data for remote surveys. The data source used for remote sensing analysis was Bing Maps Aerial (Zoom Earth, 2022), which gives access to orthographic aerial and satellite imagery worldwide. Imagery coverage varies in terms of details and depends on the area and data supplier. For this research, images supplied by Maxar in recent years were used with a spatial resolution of 50 cm and declared accuracy of 5 m. All satellite imagery used was in the WGS-84 (EPSG:4326) coordinate system.
A very important indicator of the existence of a road network, both nowadays and in ancient times, is a settlement, as travelers must usually rely on settlement for rest, food supplies, and security (Hezser 2011). In this context, two system data sources are available for the research area: the MEGA-Jordan database (2022), and the Archaeological Survey of Israel platform (2022). Both data sources make it possible to view spatial data on archaeological sites and objects. Global data repositories are another source of spatial data for archaeological sites. An important example is the Pleiades project (2022), which collects data on ancient places in the Greek and Roman world. The ancient geographical data are stored as four types of information: places, locations, names, and connections. The spatial data also contain information on the dating of sites, which allows for the analysis of data contemporary to the studied phenomena. Unfortunately, the completeness of the archaeological data in available repositories raises objections because the widely available databases are far from sufficient. Two official databases have great potential, but their use is still limited: the Survey of Israel is not complete, and it sometimes lacks any information in certain tiles (reflecting regions). This is the case with the Masada-Hebron road. As for the Jordanian website MEGA-Jordan (2022), one must apparently be a registered user to make full use of it, including accessing the information on dating and downloading shapefiles. Generally, available platforms such as the Pleiades (2022) are helpful, but mainly with regard to major archaeological sites (being the final destinations of the discovered roads), and not minor archaeological sites (sites along the course of the discovered roads, e.g., watchtowers). In summary, the use of archaeological data was limited due to the small amount of available data; at the same time, these data offer great potential for future use if the quality and accessibility of databases are improved.
It is frequently assumed that Roman Imperial roads covered earlier roads (Ben David 2009). Thus, to some extent, the network of Roman Imperial roads may be indicative of the course of pre-Roman Imperial roads. The first comprehensive attempt to collect cartographic data on the Roman road network in southern Syria was made by Smith (1874), but two modern, up-to-date publications of the cartographic data are Tabula Imperii Romani (Tsafrir et al. 1994) and Barrington's Atlas of the Greek and Roman World (Talbert 2000). Tabula Imperii Romani is prepared on 56 sheets at a scale of 1:1,000,000. The study area is described as part of the work on sheet H/I 36. In turn, Barrington's atlas was digitalized by two platforms: Mapping Past Societies (2022), led by Harvard University, and Digital Atlas of the Roman Empire (2022), maintained by the University of Gothenburg. Both spatial data sources present the same road sections. Despite the good availability of spatial data on the course of Roman imperial roads, it should be noted that they are not developed to a satisfactory scale. In the near future, major improvements are expected in this area as there are currently several projects underway to develop a coherent database of Roman roads on a detailed scale for the entire area of the Roman Empire (The Roads of the Western Roman Empire, 2022; Itiner-e, 2022; MINERVA, 2022).

Elevation data
It has been assumed in previous literature that the course of ancient pre-Roman roads may be suggested on the basis of topography (Aharoni 1979;Dorsey 1991). For past scientists, taking account of topography meant a close study of terrain obstacles in the light of cartographic sources or fieldwork experience. However, given the development of modern techniques, it is nowadays possible to move well beyond this practice, particularly by using digital elevation models (DEMs) and related GIS simulations, especially least cost paths and corridors. Generally speaking, a DEM may be described as a continuous digital representation of a terrain's surface resulting from the interpolation of the source data to numerical form (Florinsky 2012;Hawker et al. 2018;Marciak et al. 2021). Several modern methods may be used to generate DEMs, but some of the most detailed and up-to-date techniques (especially direct field surveys and LiDAR) are impossible to use in many Middle Eastern countries for political and military reasons (Arras et al. 2017 Version 2). In turn, the least cost path is a method of finding the route over a surface between two chosen points that has the least cost accumulated along the way (defined by attributes of cells being the degree of slopes in particular, but potentially also other factors, such as type of land cover or existence of settlements or blockades (Herzog 2014). The least cost path may be used to suggest, with some approximation (depending on the accuracy and level of detail of the DEM), the potential course of an ancient road (Popović and Breier 2011;Zohar and Erickson-Gini 2019) on the grounds that by avoiding the excessive effort and natural barriers, it simulates travel in a way that seems to be close to the experience of the ancients, who had less sophisticated means of travel than modern people. Likewise, the least cost corridor is a sum of accumulative costs between chosen points, and as such can be seen as a broader area through which several most probable least cost paths would have led (Pažout 2017).

Methods
The analysis assumes the combination of vector data based on the archaeological data collected from literature and the different types of raster data described in chapter 3 (Source data) in order to evaluate the potential of different types of data in the search for ancient roads. The methodology of the work can be divided into 3 basic stages: (1) data preparation, (2) data analysis, and (3) verification of the results. A detailed diagram of the research and calculation procedure is shown in the workflow in Fig. 2.
Data preparation first consisted of compiling source materials (topographic maps, aerial and satellite imagery, and DEMs). All of the collected cartographic sources needed to be georeferenced in their native coordinate system. In the case of maps from the twentieth century, the cartographic projection (Palestine Grid, EPSG: 28,191) is known; hence, the extreme coordinates of the maps were used for their georeferencing. The calibration of the nineteenth-century sources, for which the technical details of their projection are unknown (Hodson 1999), was much more complex. For this purpose, control points (temples, bridges, crossroads, etc.) were designated on all sheets of topographic maps and tied to topographic base maps. In order to fit those characteristic points from historical maps in the appropriate place, a spline transformation was used, requiring the user to have at least 10 control points per map sheet. The calibration of old topographic maps is always subject to error, which was measured for all maps used. The highest distortions, around 800 m, were noticed on the Arabia Petraea map (Musil 1906). Even older maps from the nineteenth century made by the PEF have a better accuracy of 200-300 m, which is determined by the large scale of the study of Alois Musil. The best accuracy is presented on the newer British and German military maps from the first half of the twentieth century, where the distortions are smaller (around 100 m). The last one, the Map of Israel (1956), has 300 m distortions, again because of the larger scale of the study. Road sections interpreted as ancient on old maps constituted an important part of the designated places for the analysis of satellite images and field research. When verifying the accuracy of the old maps, errors of up to several hundred meters were found on the oldest studies. Due to the wealth of information and topological clues present on the maps, we decided to use them despite their cartometric deficiencies.
All works with raster and vector data (analyses and observations of aerial and satellite data, creation and calculation of vector layers, and the graphic works) were carried out in ArcGIS Pro software (ESRI 2020).
Stage two, data analysis, began with the preparation of polyline layers representing the course of pre-Roman roads described in the literature (Strobel 1997;Kloner and Ben David 2003;Ben David 2009Ji 2019;Davidovich et al. 2022). Based on satellite/aerial images and DEM measurements, the spatial characteristics of pre-Roman roads (length, width, slopes in degrees and percentages, and difference [amplitude] of the relief) were determined. Destinations of the preserved sections of pre-Roman roads were established based on coordinates mentioned in the literature, databases of archaeological sites, and modern Fig. 2 Workflow of data processing high-resolution satellite images. Once the preserved sections of pre-Roman roads were marked out, they were divided into sections, and the slope of the terrain was calculated to capture the diversity of individual sections. Pre-Roman road sections were classified into 4 classes (0-5°, 5-15°, 15-30°, > 30°), and their percentage in total length of road was calculated. To emphasize the altitude differentiation of the roads, their relationship with the terrain was visualized in the form of profiles of road sections determined by the courses of preserved pre-Roman roads and a DEM. A key part of the GIS simulation was creating a cost surface for calculating the course of the least cost paths (LCPs) and least cost corridors (LCCs) (Kantner 2012). The raster of costs was determined based on the slope of the terrain, which was classified into 9 classes according to Jenks' "natural breaks" division. The cost of moving, therefore, increases from 1 for a flat surface to 9 for very steeply sloped terrain. Least cost paths and least cost corridors were calculated both between the start and end point of a preserved pre-Roman road and between the destination points of a pre-Roman road (places that were theoretically connected by road in the past). To create a corridor raster, it was necessary to first calculate cost distances for both connected points. At the end of the calculation process, a predetermined value was given, which indirectly determined the width of the corridor. All calculations were done separately for two variants of the DEM. The aim of the implemented procedure was to confront the GIS simulation results with the archaeologically confirmed course of pre-Roman communication routes.
The final stage involved the verification of the results of archival data and simulation results.

Results
The procedure indicated in the previous chapter required a precise determination of the course of pre-Roman roads. Using the literature, spatial data relating to 6 independent roads dating back to the pre-Roman period were collected ( Table 2). The literature was used to obtain not only information about the course of pre-Roman roads and their dating, but also about the destinations of the discovered roads. All publications contained maps, thanks to which it was possible to reconstruct the course of the roads. If the scale or quality of the illustration was not satisfactory, the course of the road was verified using satellite images.
Confirmation of the existence of the above roads in archival data is included in the "Evidence of pre-roman roads on archival data" section. All the discovered roads are located in the catchment area of the Dead Sea, which is the deepest depression in the world (Fig. 1). For this reason, all the analyzed roads face the problem of high elevations on their route. Road builders coped with the high slope of the terrain in a variety of ways, which was considered in the "Spatial characteristics of pre-roman roads" section. Finally, the use of least cost simulations in the search for ancient pre-Roman roads is scrutinized in the "Results of least cost simulations" section.

Evidence of pre-Roman roads on archival data
One of the objectives of the methodological procedure was to verify widely available archival materials in terms of the possibility of detecting pre-Roman roads on them. For this purpose, the collected cartographic sources, a mosaic of CORONA imagery and archival aerial photos, were checked for the presence of pre-Roman roads that have already been discovered.
In the case of the cartographic sources, not all of the maps covered the area of the pre-Roman roads under examination after the georeferencing procedure. This was the case with Schumacher's map sheets from 1890 and the British maps from 1946, which, although rich in content, do not cover the Dead Sea basin. However, some of the archival topographic sources even partially covering the research area are still indirectly useful for verifying the course of the discovered roads. For instance, the further course of the Wadi Zarqa Main road could be identified on both PEF's Survey of Eastern Palestine as an ancient road and on the German maps from 1918 as alte Straße leading towards Jericho (Fig. 3). Additionally, on the Survey of Western Palestine map, the Masada-Hebron road section is identified as an ancient road leading from Masada to Ein Gedi. Even without explicitly naming ancient roads, cartographic sources are a very valuable source of spatial data in that region testifying to various roads used in the past. This is the case with the Aroer Ascent, which is recorded on the map of Alois Musil from 1906. The course of Naqab Dahal leading to Buseira is also mapped on the 1:100,000 topographic maps from the years 1942 to 1956.
When considering the CORONA images in the context of the search for pre-Roman roads in the Dead Sea catchment area, it was found that it is impossible to detect remains of a small width (between 1 and 7 m). Only the widest, most evident road sections (parts of the Callirrhoe-Machaerus road and Glueck's Road) were visible in the images. On this basis, it can be concluded that the probability of identifying an unknown section of the pre-Roman road in this region using this data source is quite low.
Turning to the archival aerial photos, it should be emphasized that owing to the long-term efforts of the APAAME project, photos were available for all six pre-Roman roads. The best-documented road in terms of aerial photography is Glueck's Road-especially in the better-preserved eastern part of the Wadi Nukhayla Valley. Aerial photos perfectly show the state of preservation of the roads and their topological relationships with the landscape. In the case of attempts to identify new roads, they are of such good resolution that it is possible to use them for the correction of the courses of roads identified on satellite data.

Spatial characteristics of pre-Roman roads
The topographic characteristics of each of the pre-Roman roads were determined based on modern satellite data and the DEM (Table 3). As the measurements show, the preserved roads are very diverse, both in terms of length (from slightly over 1 km to over 20 km) and width (from 3 m to Sects. 7 m wide).
As the prepared elevation profiles of the roads indicate, the roads are also differentiated in terms of the absolute and relative heights reached (Fig. 4). Generally speaking, the mean slope calculated for pre-Roman roads varies from 7 to almost 19 degrees.
Based on the geomorphological divisions of slopes by slope (e.g., Klimaszewski 1981; Zhou et al. 2011), it was decided to divide the traced roads into sections by slope. There were 4 slope ranges (Table 4), which were related to the degree of difficulty for travelers to traverse the road. In  order to capture the slope differentiation of the discovered roads in more detail, the slope was calculated for individual segments between the vertices created during digitization.
For the 4-level classification of the inclination of the discovered roads, the percentage of a length in each class was calculated (Fig. 5). This approach makes it possible to compare the slope of road segments within a given road as well as to compare the discovered roads with each other.
An important aspect of pre-Roman roads is the issue of dealing with steep terrain, which was shown by visualizing the spatial distribution of classified roads (Fig. 6). The background of the maps is an illustrative visualization of the slope of the terrain.
Moving on to a comparative analysis of roads, only one road has manifested in its decreasing share with increasing inclination (Masada-Hebron,). The Wadi Zarqa Main and Callirhoe-Machaerus roads were also characterized by the expected slopes, with the longest class, the class of slopes from 5 to 15 degrees. It is worth noting the lack of sections above 30 degrees within the section of the Wadi Zarqa Main road. In the case of the Aroer Ascent and Naqab Dahal, the longest sections were graded from 15 to 30 degrees.
Deviations from the expected (right-skew feature in the graphs) result from the nature of individual sections, as they are only preserved fragments of longer structures. Considering the left-skew sections in the graph, these should be viewed as a more difficult segment of the longer runs: the Aroer Ascent is a steep ascent to the surface of the plateau. Glueck's Road is a route through the Wadi Nukhayla Valley, and the Naqab Dahal is a section climbing the ridge.
As part of the individual spatial analysis of slopes, changes in inclination in relation to the course of the roads were tracked. The Aroer Ascent is a short section of road with relatively high gradients resulting from a small development of the course of the road following the ridge leading to the edge of the plateau. The Callirrhoe-Machaerus road is marked out along the ridge, deftly avoiding the highest slopes and spreading the ascent over a longer distance; only near Machaerus does it use serpentines to cope with the rising altitude toward the target point. The Wadi Zarqa Main road is a slightly more favorable route in terms of overcoming terrain obstacles, where more than 80% of the road course is inclined below 15 degrees, despite initially maneuvering between small watersheds and then following the ridge leading to the vicinity of Ataruz. The longest preserved section of the Masada-Hebron road is a very favorable variant for the traveler: more than 50% of the route is inclined below 5 degrees, and the course of the route is not complicated in terms of direction changes. Naqab Dahal is initially undifferentiated; however, at the moment of overcoming the serpentines through the ridge, it consists of alternating gentle and steep sections. Glueck's Road reaches the highest slopes when overcoming terrain barriers: when the road goes down the western slopes of the valley in small serpentines, and especially in the final stage of the approach to the plateau in the eastern part.

Pre-Roman road simulation results
Simulations based on a cost surface play a very important role in the search for ancient roads for two main reasons. Firstly, they may suggest areas for in situ verification in terms of the presence of pre-Roman road remains. Secondly, they may indicate their potential course even when road remains are archaeologically elusive (e.g., due to urbanization or erosion processes). The results of simulations were prepared in two main groups: (1) between a start point and an end point of the archaeologically attested roads and (2) between two assumed destinations of a long-distance route where the attested road is only part of the route. Simulations between the start and end points of a road showed a great dependence on the resolution of the DEM (Fig. 7). As for the Aroer Ascent (see Fig. 7 (1)), both DEMs showed only general usefulness due to their resolution, as they displayed only the proper ridge of the plateau instead of the exact course of the road. The simulation results for both variants were very similar in the case of the Callirrhoe-Machaerus road (see Fig. 7 (2)). It was found that neither of them indicated the real course of the pre-Roman road: the actual road runs along the ridge, while the least cost simulations descend the ridge and follow a nearby valley. In the case of the Wadi Zarqa Main road (see Fig. 7 (3)), the simulations were different from each other and from the preserved pre-Roman road. While the real road takes a ridge west of Ataruz, the least cost path of SRTM follows a valley northwest of Ataruz, and the ASTER simulation takes a ridge north of it and descends to another valley in the northwest. The results for the Masada-Hebron road were once again different for both simulations (see Fig. 7 (4)). The least cost path of SRTM correctly indicated the course of the pre-Roman road in its initial and final fragment. The error of the SRTM simulation concerned the crossing of a vast canyon, thus shortening the length of the route at the expense of overcoming significant slopes, which the real route avoids. However, a big surprise is the result of the simulation based on ASTER data, which indicated a variant initially running north of Masada, turning west after some time through the varied terrain in terms of slopes. With regard to Naqab Dahal (see Fig. 7 (5)), the almost completely straight paths of SRTM and ASTER make it clear that such simulations do not fully work for very short stretches (1,282 m) with insufficient resolution. Both simulations incorrectly indicate the course of Glueck's Road on the western side of the valley (see Fig. 7 (6)), while the SRTM simulation correctly indicates the ridge on which the road climbs up on the eastern side of the valley.
The preserved pre-Roman roads are in fact only fragments of longer unpreserved stretches of roads that connected distant destinations in the past. In order to capture the issue of the cost rationality of the ancient routes that have been discovered, it was necessary to reconstruct the long-distance routes between the destinations mentioned in the literature for the preserved stretches. For this purpose, least cost paths and least cost corridors were determined for each of the roads in two variants based on the SRTM and ASTER models.
Simulations between destinations of pre-Roman roads showed a large variation in terms of the rationality of their determination (see Fig. 8). Rational routing was confirmed for the Aroer Ascent, and the differences between the two simulations are small (see Fig. 8 (1)). As for the Callirrhoe-Machaerus road (see Fig. 8 (2)), the simulations showed a very different route close to Machaerusthe SRTM simulation follows a ridge to the northwest of Machaerus, while the ASTER simulation uses a valley to the southwest. At the same time, the paths of the simulations become quite close in the middle of the route and near Callirrhoe. With regard to the Wadi Zarqa Main road (see Fig. 8 (3)), the course of the preserved section does not lie in the area indicated by the ASTER and SRTM simulations, which take various paths on the plateau because of its differentiated relief. A very interesting case is the situation on the Masada-Hebron road (see Fig. 8 (4)): two different GIS simulations show courses which are parallel. Remarkably, the archaeologically attested course of the pre-Roman road runs near Masada in the ASTER corridor zone and changes course at a certain distance to follow the SRTM corridor zone towards Hebron (which is also indicated by the ASTER data as a possible variant). As far as Naqab Dahal is concerned (see Fig. 8 (5)), a rational delineation of the preserved section of the route is visible within the least cost corridors, as well as in the vicinity of the ASTER least cost path. It is worth noting that in this case, the corridor simulation based on the ASTER model suggests many more possibilities, although both least cost paths have a similar course. Upon analysis of Glueck's Road (see Fig. 8 (6)), the divergence of the simulations is visible, especially close to Rujm al-ʿAbid: the SRTM simulation follows the same ridge where fragments of the ancient road are preserved, while the ASTER simulation is located more to the north. In this  (6) Glueck's Road (Wadi Nukhayla) simulation, as well as in other cases, a typical feature of the least cost corridors is visible: the corridors are very wide in the plains and plateaus, but narrow and divided into variants in the terrain with varied relief.

Discussion
On the basis of both the material collected from the subject literature and GIS analyses, this study reports a pioneering attempt to understand the features of ancient pre-Roman imperial roads discovered in the Dead Sea basin. The study also suggests some conclusions about remote sensing data and methods that may be helpful in the future search for remains of ancient pre-Roman interurban roads in the southern Levant.
As part of the research, the usefulness of many archival materials was verified in terms of the prospect of discovering new roads. Old topographic maps undoubtedly offer very promising data. Apart from explicit identifications of ancient roads, they may also be implicitly valuable by testifying to the existence of old roads, which may turn out to be pre-Roman through other means of verification. Furthermore, as has already been suggested by other researchers, topographic maps can also be useful in analyzing changes in settlement patterns (Levin et al. 2010). In contrast, CORONA imagery (Corona Atlas & Referencing System), which has proven to be so successful in the detection of pre-Roman roads in the Middle East (Ur 2003), turned out to have surprisingly little potential in our study case. The presented research also shows that the availability of a high-resolution database on Roman roads and access to comprehensive results of archaeological research is essential when attempting to capture the context of pre-Roman roads. In the coming years, major advances are expected in the case of Roman imperial road databases due to actions being undertaken in the international arena (Itiner-e, 2022). Aerial imagery, despite its high value for the verification of previously discovered  (6) Glueck's Road (Wadi Nukhayla) roads, is spatially limited and as such of very limited use in the search for new artifacts.
The use of two differently created DEMs in the analysis turned out to provide very valuable feedback. When calculating the cost area, one factor was taken into accountterrain slope, which in most models determines the course of the route to the greatest extent (Kantner 2012;Herzog 2014). In the literature on Roman imperial roads, there are examples of the use of more factors, especially intermediate archaeological sites correcting the course of least cost paths (Maniere et al. 2020;Parcero-Oubiña et al. 2017). In the case of pre-Roman roads, such practices are much more difficult to implement due to the lack of accurate data on settlement. Thus, this paper presents a comparative analysis of the preserved roads and route simulations. The best way to verify the probability of calculated road courses is to compare them with archaeologically confirmed roads (Kantner 2012).
Apart from the frequent use of least cost paths, least cost corridors were also used in this study to take other variants of simulated routes into consideration in the overall analysis of the road course. Least cost corridors offer a broader view of "motivations that might have generated real paths" (Slawisch and Wilkinson 2018). For the simulations of longer distances between final destinations (where the archaeologically attested roads covered only short stretches of the distance), the archaeologically attested roads were in most cases located at least partly in the zones marked by the least cost corridors. In contrast, due to the low resolution of the DEM, the same method turned out to be useless in the case of calculations between only the beginning and end points of the courses of the archaeologically attested roads (Fig. 9).
It must be emphasized that analysis at three levels of generalization (smaller segments of the archaeologically attested roads, entire courses of the archaeologically attested roads, and longer routes connecting destinations where the archaeologically attested roads cover only part of the distance) allowed for a broader understanding of the studied phenomenon. The most detailed scale presented  (6) Glueck's Road (Wadi Nukhayla) the great topographical diversity of pre-Roman roads. The medium scale showed that one cannot expect models with a resolution of 30 m to indicate the exact course of a route. The most general scale revealed the dominant nature of the terrain slope when determining pre-Roman roads and, in some cases, the lack of consideration of other factors such as water sources and landmarks. Certainly, a different method of determining the cost surface would have produced different simulation results. The results of the simulation should therefore be interpreted with caution, as they are strongly influenced by the applied cost function.

Conclusions
In summary, this paper has assessed the utility of different archival data sources and least cost simulations as tools serving in the detection of ancient pre-Roman imperial roads in the Dead Sea catchment area. It is evident that the use of only one source of spatial data is not sufficient in the search for new pre-Roman roads. Instead, it is necessary to gather evidence from many different sources.
Three main conclusions may be drawn from this research, each relevant for future attempts to find pre-Roman roads. First, the use of the CORONA satellite imagery, obtained from the Center for Advanced Spatial Technologies, does not yield beneficial results because of the small width of ancient pre-Roman imperial roads in this area. In the case of the Dead Sea region, archival cartographic sources fill this gap perfectly. In turn, the availability of detailed archaeological data and aerial imagery depends on a specific subarea.
Forthcoming high-resolution databases of Roman imperial roads may also offer interesting insights for research on the pre-Roman imperial road network in the future.
Second, the assessment of the results of least cost simulations shows the usefulness of the least cost corridor method for searching for variants of routes between distant destinations and the least cost path method for suggesting the most cost-effective route option on each scale of consideration. For more accurate least cost paths at a detailed scale, a DEM of a resolution higher than 30 m should be used. It would then be possible to consider variants of pre-Roman imperial roads on a detailed scale using the least cost corridor method.
Third and finally, even if the above recommendations are used in remote sensing research, there is still a strong need for archaeological research to verify remote sensing findings in the field. Cartometric errors on old maps, misinterpretations by their creators, or wrongly selected factors when determining the cost surface in the least cost simulations may result in determining incorrect courses of pre-Roman roads. What is more, field research allows the topological context of archaeologically attested roads to be captured, enables measurements on a detailed scale, and may provide information on the dating of roads on the basis of pottery finds. In this light, there is no doubt that spatial analyses, remote sensing studies, and archaeological fieldwork must constitute a coherent research effort.
Author contributions All authors contributed to the study conception and design. Material preparation, data collection and analysis were Fig. 9 Example of least cost corridor calculated for short distances (Aroer Ascent) performed by DS. The first draft of the manuscript was written by DS and MM. All figures were prepared by DS. Writing, modification and synthetic summary of data analysis, review and final paper editing were done by MM and BS. All authors commented on previous versions of the manuscript. All authors read and approved the final manuscript.
Funding This paper is part of the research project entitled "Travel and Mobility in Hellenistic and Early Roman Palestine." The project (no. 2020/38/E/HS3/00031) is financed by the National Science Centre in Poland.

Data availability Not applicable.
Code availability Not applicable.

Declarations
Competing interests The authors declare no competing interests.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.