Evaluating Subsurface Risk for Archaeological Heritage Through Ground-Penetrating Radar Surveys: The Case Study of Bisya and Salūt Archaeological Site (Sultanate of Oman)

Abstract

We present the results of a Ground-Penetrating Radar (GPR) survey conducted at the archaeological site of Bisya and Salūt (Sultanate of Oman), aimed at assessing archaeological risk associated with the planned infrastructural development of the site. The survey employed a dual-frequency GPR system with a survey rugged cart to adapt to the varying conditions of the area. The survey was designed around a scale-adaptive grid strategy, across three sectors, combining medium- and low-definition acquisitions over broader areas to identify zones with low archaeological potential, with a high-density grid near previously excavated structures. Data interpretation was integrated with Geographic Information System (GIS)-based spatial mapping, allowing the definition of a parametric risk indicator for subsurface archaeological potential derived from radar facies characterisation and point-by-point anomaly analysis along GPR profiles. Within the area of higher density, the method successfully mapped buried alignments suggestive of anthropogenic features. The results confirmed the effectiveness of GPR as a predictive tool for archaeological prospection, particularly when combined with spatial analysis. Overall, this study highlights the feasibility of incorporating non-invasive methods into heritage protection strategies, contributing to the sustainable development and planning of archaeological landscapes.

1. Introduction

Geophysical methods stand out for their ability to provide rapid, extensive and detailed data on near-surface anomalies with a totally non-invasive approach [1,2,3,4,5]. They are particularly suitable for large-area assessments, where the balance between time-efficiency, resolution and minimal impact is essential. These methods can reveal the presence of structures of potential archaeological interest in the near-surface soil horizons and therefore are of paramount importance for multiple goals: (i) supporting the design of archaeological excavation; (ii) mitigating potential impacts on cultural heritage resources of infrastructure planning; (iii) integrating heritage protection into broader territorial development strategies. Among many other techniques of geophysical prospecting, Ground-Penetrating Radar (GPR) is a flexible, quick and effective tool to assess the potential for buried archaeological features in the case of bare surfaces [6,7,8,9,10,11,12,13,14,15], where it is possible to plan and perform surveys at different degrees of resolution. As the absence of obstacles and vegetation is a key to the efficiency of GPR application, arid and semiarid regions—almost lacking plant cover [16]—offer good opportunities for this kind of investigation.

In such a context, a successful geophysical survey has been performed inside the Bisya and Salūt archaeological site in central Sultanate of Oman. The work was promoted by local authorities that required non-invasive investigations to assess the archaeological potential of the area surrounding the Bronze to Iron Age remains, notably including the Iron Age Salūt citadel and settlement (identified as Husn Salūt and Qaryat Salūt, Figure 1), planned to host touristic infrastructures.

The main goal of the GPR survey at Salūt was to assess site-related archaeological risk, in the context of the expansion of touristic infrastructures (namely an access road and a drop-off area) within the Bisya and Salūt archaeological site. In the jargon of the civil engineering community, this risk is related to archaeological potential, namely the likelihood of encountering archaeological remains in the subsurface and the probability that infrastructural works may interfere with them.

Evaluating this potential is essential to reconcile the archaeological park development with the crucial need to preserve its archaeological components. In the present case study, the assessment focused on the area designed for the construction of an infra-site visitors’ road network and a drop-off area. Therefore, the ultimate goal was to support the assessment of potential damages to archaeological remains caused by infrastructural works and possibly suggest a road layout which reduces the archaeological risk. Moreover, GPR is also a well-established technique as a predictive method for pinpointing promising excavation sites. In fact, it allows identifying subtle variations in subsurface horizons when sufficient contrast in the relevant dielectric properties of soils exists and is highly versatile on gently undulating or even slightly rugged terrain. In addition, GPR has a relatively rapid deployment, producing preliminary results directly on site. So, a secondary goal of the survey was also the realisation of a high-density survey in a limited area, to support future archaeological excavation.

However, GPR survey in the specific logistical condition at Salūt required a scale-adaptive survey strategy, with different scales of resolution, but it finally demonstrated the possibility to adapt a GPR-based approach to preventive archaeology in complex topographic and archaeological contexts.

2. GPR Applications to Arabian Archaeology

GPR application has proven particularly effective in sand dune environments in different areas of the world, where it can potentially reveal sedimentary structures to a depth of more than 10 m below the ground surface [17].

In recent years, it has consistently become increasingly frequent in Southeast Arabian archaeology. At the Iron age site of Muweilah in the Emirate of Sharjah, it was implemented to integrate previous results obtained with a gradiometer [18], which only reached the depth of 1 m and generally does not detect the same range of materials [19]. At the multi-period site of Saruq al Hadid (UAE), a combination of GPR, excavation and radiometric data were used to create a model of the sand-body of a stratified dune [20]. GPR was also implemented at the prominent Late Iron Age site of Mleiha (inland Sharjah, UEA), where soil conditions similar to those at Salūt attenuate the signal [21,22].

Sometimes, the results were not satisfactory, due to the challenging environment. While settlement areas and monumental funerary structures were identified with sufficient certainty, isolated graves were less clearly located, and humidity retention by plant roots also played a role in blurring the results. In several occasions, GPR has been used in combination with magnetometry as, for example, at al Madam near Mleiha [23,24], in the oasis of Buraimi near the UAE/Oman border [25], and in the Sultanate of Oman, at the Early Bronze Age site of al-Ghubra near Bahla [26], less than 30 km north of Salūt as the crow flies.

Preliminary GPR survey of a limited area in the Salūt plain was undertaken between 2011 and 2013 by geophysicists from Swansea University; however, it remained only sketchily published, together with the results obtained at nearby Early Bronze Age sites along Wadi Bahla, close to where it reached the town of Bisya [27,28].

3. Archaeological Background

The Archaeological Sites of Bisya and Salūt encloses more than 60 hectares of flat land located just north of the town of Bisya, in central Oman (Ad Dakhiliyah Governorate), as well as part of the rocky hills that border the plain to the northeast. These reliefs are indicated in the geological maps as being part of Jabal Hammah (Geologic Map of Oman, 1:250,000, sheet NF4007 e NIZWA), although locals currently refer to them as Jabal Salūt.

Fifteen years of extensive archaeological research by the Italian Mission To Oman of the University of Pisa, supported in the recent years by a collaboration with the Department of Earth Sciences of the University of Milan, shed light on the nature of several sites with different chronologies, e.g., [29,30,31], spanning the Early Bronze Age to the Late Iron Age (roughly 2500 BC–100 AD in this specific case), with additional evidence for repeated occupation after the advent of Islam [32,33].

The inclusion of all these sites, which represent different types of archaeological evidence (residential, collective, cultural, and funerary sites), makes the archaeological site a significant and representative legacy of the variegated archaeological landscape of the region, which also comprises a rich rock art record [33].

In the plain where the new tourist infrastructures are going to be realised, three main poles of attraction coincide with three main sites: Husn Salūt and Qaryat Salūt; Salūt-ST1; and Salūt-SLP. The latter is a Late Iron Age (c. 150 BC–100 AD) necropolis, 39 underground graves of which have so far been investigated [34], and stands in a place where the impact of infrastructural development should be likely limited to the setting of a footpath track (Figure 1, green dashed line over white background). The Iron Age settlement of Husn and Qaryat Salūt represents the core of the Site, and Husn Salūt is the key site that sparked the interest of the local authorities and led to the start of the research project in 2004. Husn Salūt and Qaryat Salūt are two labels chosen by the excavators to indicate the elevated, fortified, and mostly communal part of the site (Husn) and the surrounding, partly terraced and partly built on the plain, residential area (Qaryat). The site was established around 1300–1250 BC, if not earlier, and continued to be occupied until the 1st century AD at the least [31,35]. In the western part of the Site, Salūt -ST1 is an Early Bronze Age “tower” site, the typical monument for this period in south-east Arabia. ST1 is characterised by an impressive water management system, which comprises substantial channels and ditches cut through the cemented substratum of the plain [36]. Its occupation can be dated between 2500 BC and 2000 BC ca. [37].

Given their extension, none of these sites has been explored in its entirety. At ST1, the complete outline of a ditch that surrounds the central tower was revealed, but other channels were only partially exposed and their plan extension remains unknown, though certainly crossing part of the archaeological site.

At the SLP necropolis, several low mounds suggest the presence of other burials, mainly to the south and west of the excavated core area, but similar traces are lacking toward the hills, hidden by thicker alluvium.

Finally, Qaryat Salūt certainly occupies the whole hill that also hosts Husn Salūt, but its extension onto the surrounding plain remains uncertain, except for the northern limit marked by a substantial stone wall. A first set of test trenches revealed the presence of built structures all around the hill. They extend as far as at least 60 m from its footings to the east, where the construction of one of the inner tourist roads is envisaged. Moreover, the preliminary investigations showed that in several instances the surviving structures are buried below only a few tens of centimetres below the current surface.

In light of these dense resources, it is evident that any building work in the area would impact relevant archaeological features, which made the GPR survey of the utmost importance.

4. Geomorphological Setting

The study area is located near the confluence between Wadi Sayfam (to the east) and Wadi Bahla (to the west). Available geological maps [38] indicate that the bedrock of the region consists of a series of sedimentary and igneous /metamorphic rocks. To the north, the study area is framed by the Jabal Al-Akhdar Mts., where tectonized harzburgite and intrusive peridotite and gabbro reliefs of the Mid-Late Cretaceous Samail Ophiolite outcrop; to the west and south the area is bounded by outcrops of Permian to Cretaceous limestone and radiolarite formations. A system of coalescent, gravelly alluvial fans stretches out from the northern ophiolitic hills and runs parallel to Wadi Bahla, merging with the plain at the eastern margin of Wadi Sayfam. The alluvial plain of Wadi Sayfam and Wadi Bahla—including the plain surrounding the site of Salut—is covered by fine sediments consisting of wind-blown silt, trapped by bush grassland developed when climate was substantially wetter than today [39] and partially reworked by fluvial process (Khabra Formation). Today this formation does not include any aquifer, reaches a mean depth of about 1.5 m below the ground surface and overlays consolidated fluvial sediments. Test trenches excavated in the area show a stratigraphy including an uppermost layer of loose, stone- and potsherd-rich, sandy loam deposit, likely formed by the deflation of older sediments. This layer is followed by a weakly developed soil horizon characterised by poorly aggregated, pale brown sandy loam containing small charcoal fragments and scattered pottery fragments that cover a Bk soil horizon (calcrete) developed on sandy and silty alluvial sediments [36,39] (Figure 2).

5. Method

In this section, after briefly introducing the working principle of the GPR, the surveys strategies specifically developed for the Salut site are described, together with the survey designs and data management and treatment strategies.

5.1. GPR Method

Ground-Penetrating Radar operates by transmitting high-frequency electromagnetic (EM) waves into the subsurface. EM waves emitted from a transmitting antenna propagate through the ground and are partially reflected back to the surface, when they encounter boundaries characterised by contrasts in electromagnetic properties, such as changes in soil composition, the presence of man-made structures (e.g., walls, foundations, pavements, voids, stony materials and other), or variations in moisture content. The reflected EM waves are re-collected as signals by a receiver antenna in terms of amplitude, phase and two-way travel times (in nanoseconds) of EM wavelet arrival. By dragging the antennas along the topographic surface, GPR yields a nearly continuous 2D image of the reflectivity through the subsoil along straight profiles and at increasing depth (radar profiles).

The main properties affecting EM wave propagation and signal absorption in non-magnetic earth materials are dielectric permittivity and electrical conductivity. These parameters influence EM wave velocity, signal attenuation and reflection amplitude: dielectric permittivity is mostly sensitive to moisture and air content, while electrical conductivity is mostly sensitive to pore fluid chemistry, stiffness (loose or cemented deposits, stony material) and lithology, such as granular, sand-like vs. cohesive, clay-like materials.

Typically, radar frequency band in GPR application varies from 50 MHz to 2.0 GHz, and the choice of antenna frequency is a crucial aspect in GPR surveys, as it influences both the depth and the resolving power of the investigation. Higher probing frequencies (e.g., peak central frequency of 800 MHz or more) provide better resolution GPR data, allowing for the potential detection of small and shallow features in the ground, but they result in limited penetration depth, due to strong EM attenuation in the medium. Conversely, low to medium frequencies (e.g., between 100 MHz and 500 MHz) allow deeper sounding into the ground, but at the expense of reduced spatial resolution. Therefore, the selection of an appropriate operating frequency involves a balance between the need for adequate resolution to identify targets of interest and the ability to penetrate to the required depth, depending on the specific geological and archaeological context. The conditions found at the ground surface, as a function of the presence of rugged terrain, also influence the ability to receive and correctly interpret coherent reflections from the subsoil.

In the unique setting of the Salūt archaeological site, near-surface conditions are shaped by a low-moisture environment composed of loose sandy loam deposits, overlain by a weakly developed soil horizon characterised by poorly aggregated, pale brown sandy loam (Figure 2). The contrast between these loose, unconsolidated sediments and denser, more resistive, stony materials creates a potential dielectric and electrical contrast, which is favourable for the reflection of electromagnetic waves from buried archaeological features, as detected through GPR imaging.

5.2. GPR Strategies Followed at Salūt

The selection of the GPR survey areas was guided not only by the ultimate goals, but also by the results of a preliminary archaeological field assessment, primarily focused on the distribution of surface findings (such as scattered pottery sherds) as indicators of potential subsurface archaeological features. In addition to archaeological data, the layout of the planned infrastructures was considered to design the GPR survey, with particular emphasis on the zones designated for the construction of the drop-off area (Figure 1, in light green). These factors helped prioritise locations where the likelihood of archaeological remains was highest and where construction activities were expected to have the greatest impact on the subsurface.

The survey setup was specifically adapted based on the general targets, the required spatial resolution, and local logistical factors like surface type and area size. For each area, a scale-adaptive grid strategy was adopted, with the density and GPR profiles layout being adapted to local surface conditions and targets.

The principal GPR exploration was performed across the main project exit-way, by a wide-scale, horizontal mapping of EM reflectivity along three wide areas (namely sectors 1a, 2 and 3 in Figure 1 and Figure 3) to reveal shallow built features (down to 1 m below ground surface) that were vulnerable to road works and to identify reflection-free corridors with low potential of archaeological findings.

A second exploration was performed by acquiring high-density subsurface imaging in the proximity of previously excavated ancient structures (sector 1b, in Figure 1 and Figure 3), to reveal further, shallow, built features.

The design of the GPR survey required a non-standard approach, due to the spatial extent of the survey area and the varying degree of surface accessibility. In particular, a preliminary inspection showed mainly horizontal areas of vehicle transit, even with deep tracks and featuring deep loose sand, and areas with scattered and clustered vegetation and bushes. Therefore, two types of GPR surveys were conducted (Figure 1), each with different grid densities and data coverage:

• Coarse- and middle-resolution GPR surveys were conducted where surface conditions were less favourable (Figure 1, yellow dashed rectangles and red dashed rectangles, respectively), using widely spaced GPR profiles, aiming to statistically assess the presence (or absence) of significant subsurface EM anomalies across the area. This approach was not intended to obtain a high-density subsurface imaging; instead, it was designed to identify broad reflection-free corridors with low archaeological potential, particularly within the first metre below the ground surface. The coarse grid spacing allowed efficient coverage of large areas in a relatively small amount of time, while still enabling the detection of major anomalies without fine-scale spatial detail.
• Fine-resolution GPR survey was conducted where surface conditions were more favourable, allowing for improved data acquisition and resolution (Figure 1, blue dashed rectangle), by employing closely spaced GPR profiles to achieve detailed 3D subsurface imaging. This configuration was specifically designed to map small-scale EM anomalies with high precision, enabling the identification of archaeological features. The dense acquisition grid ensured maximum resolution and interpretative accuracy in areas where the presence of archaeological remains was suspected based on previous excavation.

5.3. Survey Systems and Design

To optimise the trade-off between resolution and penetration depth, GPR survey at Salut Castle was conducted using a GSSI SIR 4000 system equipped with a ground-coupled, dual-frequency antenna operating simultaneously at the central frequencies of 300 MHz and 800 MHz (Figure 4A), respectively, with 20 ns and 63 ns acquisition window. With this configuration, the 800 MHz frequency channel enables to acquire fine-resolution data in the uppermost layers of the subsurface, which is ideal for detecting small or near-surface archaeological features. On the other side, the 300 MHz frequency channel achieves deeper penetration, allowing the identification of larger or more deeply buried structures.

To ensure accurate and spatially referenced data acquisition, a detailed local survey grid was established before the GPR survey, using a high-precision Leica total station for the accurate positioning of GPR profiles (Figure 4B). GPR profiles were generally collected along straight traverses, and the equipment was manually dragged over a rugged topographic surface at an average speed of less than 1 m/s to minimise noise due to antennas’ roll and pitch, with an horizontal sampling equal to 1 scan per 2 cm. Areas with dense vegetation or surface debris were necessarily avoided, while flat and accessible zones were preferentially selected for data acquisition. On the other hand, in sectors where deep cuts caused by vehicle tyres were present, data quality was necessarily poor due to uneven contact between the antenna and the ground surface.

GPR surveys conducted across the three sectors (Figure 1) focused on two main targets:

• Archaeological risk assessment along the planned project exit route:

  ○

  Sector 1a (Figure 3 and

  Table 1. Summary of parameters of GPR data acquisition.

  Sector	Area (m2)	Profile Spacing
  1a	8000	2 m; orthogonal grid
  1b	700	0.5 m; orthogonal grid
  2 and 3	12,500	10 m; parallel profiles

  ): this area covers approximately 8000 m² and is located in the eastern sector of the site, corresponding to the planned exit route connecting to the main access road. A medium-resolution survey was carried out using a 2D acquisition grid composed of orthogonal GPR profiles spaced 2 m apart in both directions (2 m × 2 m grid). The total length of GPR profiles acquired exceeded 7500 m.

  ○

  Sectors 2 and 3 (Figure 3 and Table 1): these areas cover approximately 12,500 m² and they correspond to the south-eastern and north-western portions of the proposed exit route, respectively. A coarse-resolution survey was performed in these widespread areas, with GPR profiles spaced on average 10 m apart. The total length of acquired profiles in these sectors exceeded 1250 m.

• Subsurface imaging near the 2019 excavation area:

  ○

  Sector 1b (Figure 3 and Table 1): this area, covering approximately 700 m², is located on the eastern flank of the Husn and adjacent to the archaeological excavation conducted in 2019. A high-density GPR survey was performed using a 2D acquisition grid of orthogonal profiles, spaced 0.5 m apart and with a total length of profiles exceeding 4000 m.

5.4. Data Processing and Management

GPR data were processed using GSSI RADAN 7 software to enhance the recognition of reflected signals of archaeological interest by reducing background scattering and noise. The reference data processing workflow for each GPR profile consisted of:

(1)

Time-zero setting, to set the first-event recognition corresponding to the reflection from the ground surface.

(2)

Gain amplification, to enhance the late arrival reflection event.

(3)

average dielectric constant evaluation through Kirckhoff type hyperbola best-fitting.

(4)

Band-pass (200 Mhz high pass and 900 Mhz low-pass) and moving-average (back-ground removal) filtering.

(5)

Three-dimensional assemblage of collected GPR profiles in a grid format and reflectivity slicing (only for sector 1b).

Regarding velocity estimation and time-to-depth conversion, the latter was based on the assumption of a homogeneous shallow subsurface and on the average velocity derived from best-fitting the most clearly expressed hyperbolic anomalies in the GPR dataset. This approach yielded a dielectric constant (ε_r) of 5, corresponding to a velocity of approximately 0.13 m/ns, which was then applied for depth conversion across the survey area. Due to noise in the GPR profiles caused by surface roughness, a few less distinct hyperbolas produced dielectric constants up to 7. The poor quality of these noisy hyperbolas and therefore of the corresponding fitting parameters prevented us from considering the results as reliable and useful for time-to-depth conversion. Nevertheless, the variability in the fitted values of dielectric constant can be ultimately used to obtain a rough estimate of the relative uncertainty on the depths, which is of the order of 15%.

Once the standard data processing workflow was applied, different methodologies were implemented for the subsequent data processing steps, depending on the resolution and objectives of each survey type.

For the low- and medium-resolution survey areas (see Section 5.1), each profile was examined to detect anomalous subsurface regions based on variations in reflectivity and diffraction patterns at both working frequencies, here referred to as radar facies. For the purpose of this study, we define a radar facies as the combination of reflection pattern features plausibly generated by sedimentary deposits hosting man-made structures or modified by anthropic activity. Such anomalies are qualitatively recognisable in GPR profiles through variations in reflection amplitude and density and through the presence of diffractions likely determined by the structural and textural properties of the subsurface.

Two main radar facies were identified (Figure 5):

• Type A radar facies: complex and chaotic reflections, with non-fully expressed hyperbola diffractions.
• Type B radar facies: coherent and well-defined anomalies, including isolated or clustered hyperbolic diffractions.

After this interpretation, GPR anomalies were georeferenced and mapped onto the project’s base plan as point anomalies. To assess the archaeological risk along the planned route of the perimeter road adjacent to the site, a map of the areal density of potential subsurface structures was generated using Geographical Information System (GIS) tools, with the following procedure. For each cell (whose side-length is equal to 0.25 m), the number of single GPR point anomalies that fall within a neighbourhood (with a radius of 6 m, equal to 3 times the GPR profile spacings) around the cell was computed; the obtained values ranges from 0.00 points/m² to 0.30 points/m² were clustered in five categories for a simple representation of the archaeological risk, adopting 0.15 points/m² for medium risk and 0.25 points/m² for high risk (Figure 6).

For the high-density dataset (see Section 5.2), processed GPR profiles were used to construct 3D models of subsurface reflectivity indicating the raw energy reflected from an object or a layer. From these models, horizontal amplitude slices (reflectivity maps) were extracted at various depths, down to the maximum exploration depth obtained at low-frequency (Figure 7).

These maps enabled the identification and tracking of linear alignments and clusters of radar anomalies with a high degree of spatial precision. This detailed approach facilitated the interpretation of small-scale archaeological features and their spatial relationships.

6. Results and Discussion

6.1. Archaeological Risk Assessment Across the Project Exit-Way

In sector 1a (Figure 6), many low-relief anomalies and a small number of more coherent anomalies were identified in the near-surface layer less than 1.5 m below ground level. The horizontal density map, derived from GPR data acquired using a 2 m × 2 m grid, indicates that anomalies are primarily clustered within the central portion of the sector. Isolated anomalies are also present, but their lack of continuity across adjacent GPR profiles limited a more detailed interpretation. In the western part, the presence of a restricted-access area (due to dense vegetation, red tilted lines pattern in Figure 6) prevents further evaluation toward the Husn.

The results obtained in sector 1a highlighted that the proposed road layout, particularly the drop-off area (Figure 6, black-dashed line), potentially intersects zones characterised by a high density of subsurface anomalies. A revision of the original design has been, therefore, proposed (Figure 6, green-dashed line), suggesting a more precautionary relocation of both the access road segment from the site entrance and the drop-off area itself.

Concerning the detected GPR anomalies, the profile density in sector 1a is not sufficient to accurately define detailed plan-view geometries or to infer the precise nature of the materials responsible for the EM reflections. These aspects will necessitate verification through direct investigation, specifically the excavation of archaeological trenches, in a subsequent phase.

In sectors 2 and 3, low-density surveys revealed only a limited number of near-surface GPR anomalies, which appear sparse and isolated, with no identifiable clustering (Figure 3). For these GPR anomalies we report only the details already provided in Figure 3, where yellow dots are used to locate complex anomalies (medium to low reflectivity) and light blue dots are used when no noticeable anomaly was detected. In sector 3, some linear alignments emerge in the GPR profiles collected from the northern area. Nevertheless, the overall characteristics of the subsurface inferred from the GPR data across these sectors suggest a very low archaeological risk related to the road segment, due to the absence of continuous, non-natural structures, so that the proposed layout in these sectors can be confirmed without the need for further modifications.

6.2. Subsurface Imaging in the Proximity of Previous Excavation

In sector 1b, adjacent to the area excavated in 2019, the high-resolution 3D survey allowed for the identification and mapping of numerous anomalies using reflectivity maps (Figure 7). These features are located approximately within 1 m below the ground surface and appear as high-reflectivity zones, including both localised and linear alignments.

Particularly, the northern portion of the sector exhibits significantly higher reflectivity compared to the southern part. Clearly defined, orthogonal, and intersecting features are visible, and in some cases, these extend beyond the limits of the survey grid, suggesting that the associated structures likely continue outside the investigated area. The data supports the presence of a densely built-up sector of the settlement in this area. This confirms that the Iron Age village (Qaryat Salūt) surrounded the small hill above which the fortified part of the site (Husn Salūt) was erected on at least three sides (east, north, and west). The only possible exception is the southern area, potentially due to the accentuated steepness of the slope that would have complicated the communication between the terraced part of the settlement and the one built on the plain.

7. Conclusions

The GPR survey conducted at the archaeological site of Bisya and Salūt underscores the utility of applying non-invasive geophysical methods in archaeological risk assessment and subsurface imaging, also highlighting the need for accurate survey design and flexibility in the scientific approach to obtain good quality data adequate for the ultimate practical goals.

The survey was tailored to both logistical constraints and the archaeological priorities found in the context of an infrastructure design, e.g., the construction of an infra-site visitors’ road network. It provides preliminary insights into the spatial distribution of near-surface anomalies likely associated with buried features, which are extremely useful for the planning of sustainable management and development of the site, while they can also inform further excavation strategies.

We demonstrated the versatility of the GPR technique: when combined with GIS- based mapping capabilities and precise georeferencing of data, it allows for the development of an operational and user-friendly protocol for the interpretation of GPR results. This protocol enabled the generation and spatial analysis of an indicator parameter of subsurface archaeological risk across relatively large areas, through the definition of characteristic radar facies and the detailed analysis of radar signatures along each GPR profile, thus producing a rapid and interpretable mapping of areas with a higher likelihood of archaeological features. In a future perspective, the proposed definition for radar facies could be further enhanced through attribute-based analyses and machine learning approaches, which may allow for a more systematic and automated characterisation of radar facies and anomaly patterns in archaeological prospection [40].

In sector 1a, medium-resolution data highlighted zones of high subsurface reflectivity, suggesting a potential concentration of archaeological remains. This result led to a proposed revision of the road network layout to minimise impact on potentially significant areas. In contrast, sectors 2 and 3 revealed only sparse and isolated anomalies, supporting the viability of planned infrastructure development in these areas without further archaeological mitigation.

In the high-resolution survey performed in sector 1b, adjacent to previously excavated areas, the results confirmed the high potential of the GPR method to map archaeological features at a very fine scale, particularly when results can be integrated with data from previous, nearby excavations.

The level of detail in the GPR results at Salūt necessarily varies from one area to another but, as a significant methodological outcome, the results at all three areas highlighted how the versatility of the GPR method, when combined with GIS-based mapping capabilities and precise archaeological background data, enables the development of an operational protocol for the interpretation of GPR results, demonstrating the usefulness of GPR for supporting archaeological risk mapping in such contexts

This approach not only enhances the reproducibility and accessibility of geophysical interpretations but also represents a replicable model for archaeological risk assessment and site management in similarly complex environments, also reinforcing the necessity of the early use of non-invasive diagnostic tools for the sustainable management and enhancement of cultural heritage sites.