Making the Invisible Visible: The Applicability and Potential of Non-Invasive Methods in Pastoral Mountain Landscapes—New Results from Aerial Surveys and Geophysical Prospection at Shielings Across Møre and Romsdal, Norway

Abstract

Shielings are seasonal settlements found in upland pastures across Scandinavia and the North Atlantic. New investigations in the county of Møre and Romsdal, Norway, demonstrate the existence of this transhumant system by the Viking Age and Early Middle Ages. Sub-terranean features in these pastoral mountain landscapes have been identified by remote sensing technologies, but non-invasive methods still face challenges in terms of practical applicability and in confirming the presence of archaeological sites. Generally, aerial surveys, such as LiDAR and image-based modelling, excel in documenting visual landscapes and may enhance detection of low-visibility features. Thermography may also detect shallow subsurface features but is limited by solar conditions and vegetation. Magnetic methods face challenges due to the heterogeneous moraine geology. Ground-penetrating radar has yielded better results but is highly impractical and inefficient in these remote and rough landscapes. Systematic soil coring or test-pitting remain the most reliable options for detecting these faint sites, yet non-invasive methods may offer a better understanding of the archaeological contexts—between the initial survey and the final excavation. Altogether, the study highlights the dependency on landscape, soil, and vegetation, emphasising the need to consider each method’s possibilities and limitations based on site environments and conditions.

1. Introduction

The historical division between cultivated infields and uncultivated mountains and outlands (Norw. utmark) has not only shaped the traditional Norwegian farm landscape but continues to influence cultural heritage management and current methodologies employed in archaeological surveys and prospection. Infields and central agricultural areas are usually surveyed through mechanical topsoil stripping [1] and, more recently, motorised ground-penetrating radar [2,3]. The outlands, however, are mainly investigated through visual inspection, often aided by aerial photos and airborne LiDAR data readily available online [4,5], as most sites and features in these more ‘natural’ environments are likely to remain visible above ground.

The typical shieling landscape, however, is situated somewhere in between infields and outlands. Traditionally, until the mid-20th century, the sites were part of a transhumant system with seasonal movements of livestock. The cattle were kept in the mountains and forests throughout the summer—mainly in order to save the grassland at the permanent farm for the collection of winter fodder [6] (p. 2). Gradually, these seasonal and pastoral sites in the outlands grew into something reminiscent of small farms, with multiple building types and often with cleared, enclosed meadows to further increase winter fodder production [7]. Some of the sites also grew into permanent farm settlements, and the boundary between farm and shieling is often blurred and interchangeable [7,8].

Despite being genuine agro-pastoral landscapes, continuously transformed by long-term occupation, archaeological surveys at shieling are still focused on documenting visual features. Hand-dug test pits often reveal archaeological deposits, with radiocarbon dates spanning the Iron Age and Medieval period [7]. However, we still know very little about the potential for finding subterranean features, such as house foundations, fences, post-hole pits, and fireplaces. Due to low development pressure, archaeological excavations are rare [9,10], and mechanical excavators are seldom—and with varying results—brought to these remote mountain sites. Besides, historical shielings are often highly valued heritage sites for in situ preservation and recreational purposes. Thus, the physical structure and the activities carried out in these early pastoral environments have remained in a blind spot [11] (pp. 54–55).

The aim of this study is to investigate the potential for remote sensing and non-invasive methods in this type of valuable and vulnerable landscape [12] as a way of mapping cultural heritage sites and features with low or no visibility above ground. The project covers both aerial surveys employing unmanned aerial vehicle (UAV)-based models, such as LiDAR scanning, image-based modelling (IBM), and thermal photography, and geophysical prospection by means of ground penetrating radar (GPR), gradiometer, and topsoil magnetic susceptibility mapping. What are the possibilities and limitations of employing these methods in remote shieling landscapes, both in terms of archaeological results and practical applicability across diverse environments and conditions, and how can these technologies be implemented in current cultural heritage management practices?

2. Materials and Methods

We have examined a selection of five shielings across the county of Møre and Romsdal, Norway, with varying topography, geology, and vegetation cover (Figure 1). Across the region, metamorphic gneiss is the dominating parent material. Drift geology varies, but glacial moraine deposits dominate most of the shieling sites [13]. Furthermore, none of the selected sites are densely forested, and for purely practical reasons, all the sites are located close to nearby roads. The main criterion is that they all contain archaeological remains dated between the Migration Period and the High Middle Ages (AD 400–1350), mainly documented by test pitting and small keyhole excavations.

Non-invasive investigations were conducted in several rounds during the summer of 2022 (phase 1). In early June, with the vegetation still being low, we performed a three-day campaign using UAV-based survey methods (LiDAR, IBM, and thermography). Magnetic surveys (gradiometer and topsoil magnetic susceptibility) were mainly carried out in mid-to-late June, whereas the GPR surveys were conducted in early July and September. In cases where timing may affect dataset comparisons, this will be further detailed below.

To assess the practicality and efficiency of applying the varying methods within a developer-led survey context, we established a self-imposed maximum time limit of half a day for fieldwork with each approach. The extent of each survey method is thus a measure of applicability. The varying results produced by the different technologies are compared based on visual interpretation and their correlation to already established background knowledge. Despite technological advancements and ongoing efforts in data fusion—such as the multi-instrument integration of heterogeneous datasets—visual interpretation relying on human expertise remains a cornerstone in archaeology and cultural heritage research [14,15]. To exhibit various degrees of certainty, the interpreted features are marked as either definite, probable, or possible.

More systematic ground truthing was performed in 2023 (phase 2), with excavation permits granted from the Norwegian Directorate of Cultural Heritage. All the sites (Figure 2) were revisited and investigated through soil coring and small-scale manual trenching of a representative sample of potential subterranean archaeological features.

2.1. Light Detection and Ranging (LiDAR)

LiDAR is an active remote sensing technique that involves the use of a laser scanning range finder to produce height data in the form of a point cloud. By emitting electromagnetic energy in the form of a laser pulse, the LiDAR sensor records the time it takes for the laser pulse to hit a surface and reflect back to the sensor. For airborne LiDAR systems, a navigation unit consisting of a GNSS receiver and an inertial movement unit (IMU) provides accurate information on the position and orientation of the sensor, allowing us to calculate coordinates for every reflected pulse. The sensor utilises a rotating mirror to distribute the laser pulses in a line perpendicular to the aircraft’s trajectory. Depending on the sensor’s specifications, it can record a given number of signal returns per laser pulse (usually three or five returns for UAV systems), allowing for multiple data points for each pulse in areas of dense vegetation. This makes the LiDAR system ideal for topographical mapping in forested areas, producing highly detailed 3D models of the forest floor beneath the canopy [16,17]. The LiDAR technology has revolutionised the field of archaeological surveying, leading to groundbreaking discoveries in areas where traditional manual field surveys have been a challenge [18,19]. The ever-growing repository of publicly available LiDAR data has also led to increased implementation of the method as a tool for the ordinary day-to-day cultural heritage management practice in many parts of the world [20,21,22].

The LiDAR data for this project was collected from a Riegl MiniVUX 3 system mounted on a DJI Matrice 600 Pro drone. The system utilises an Applanix APX20 IMU/GNSS unit for positioning and trajectory correction. This correction used base data from the CPOS GNSS reference system (RTK) and was undertaken during post-processing with the Applanix PosPAC UAV (v. 8.8) software.

The areas were scanned at an altitude of 107 m above ground level (m.a.g.l.) and a cruising speed of 8 m/s, with the scanner at full power, flying a double grid. This approach produced datasets with an average point spacing of 0.117 m across all sites after vegetation filtering and edge trimming of the point cloud. This equals an average point density of 77 points per m².

The LiDAR data were processed and classified in Riegl’s RiProcess (v.1.9.3.6) software and exported as .las files with both elevation and laser intensity data. Further analysis and visualisation were carried out in ArcGIS Pro (v. 3.4) after converting the .las files to raster images with a 0.10 m cell size. Hillshade-, multidirectional hillshade-, aspect-, slope-, and local relief models are some of the most common techniques for visualising LiDAR data [23,24]. All these techniques were applied to the acquired data, with individual tweaking of z-values, gamma enhancement, and histogram adjustments to reveal the minute details in each data set (

Table 1. Statistics for UAV based data acquisition.

Site and Method	Area (km2)	Aligned Images	Points/ Vertices (Millions)	Point Spacing (m)	Point Density	Flight Pattern	GSD (cm)
Klovset
LiDAR	1.070	-	30.31	0.111	81.2	Cross grid	-
IBM	0.053	1047	102.49	0.021	2178	Lines	13.71
Thermal	0.047	962	7.00	0.082	93.2	Lines	5.00
Setersetran
LiDAR	1.893	-	19.58	0.109	84.2	Cross grid	-
IBM	0.020	1329	86.06	0.015	4303	Lines	6.86
Thermal	0.018	1394	7.28	0.050	440.4	Lines	3.46
Myrset
LiDAR	1.449	-	29.71	0.110	82.6	Cross grid	-
IBM	0.023	1151	94.06	0.015	4090	Lines	8.22
Thermal	0.022	1122	6.67	0.057	267.3	Lines	3.67
Søstølen
LiDAR	0.678	-	39.86	0.101	98.0	Cross grid	-
IBM	0.050	906	180.49	0.016	3609	Lines	13.71
Thermal	0.032	721	8.47	0.061	361.1	Lines	3.19
Rangsetra
LiDAR	1.635	-	6.07	0.157	40.6	Lines	-
IBM	0.032	582	63.18	0.022	1973	Lines	10.97
Thermal	0.015	543	3.42	0.066	153.2	Lines	4.05

).

2.2. Image-Based Modelling (IBM)

IBM is a computer graphics and vision technique for creating 3D models from overlapping 2D images. The IBM algorithms analyse the visual information present in the images and infer the geometry and appearance of the underlying 3D structure [25]. The technique originated from early photogrammetry and computer vision research in the 1990s, when advancements in structure-from-motion (SfM) and multi-view stereo techniques enabled more robust 3D reconstructions. The boundaries of realism in virtual environments were pushed in the 2000s, with increasingly photorealistic rendering and texture mapping.

Since the 2010s, the integration of machine learning has steadily improved the accuracy and efficiency of the technique [26]. During the same decade, IBM became a crucial tool in archaeology and cultural heritage documentation, allowing for the creation of high-resolution 3D models of artefacts, structures, and landscapes. The technique has democratised digital preservation by making it more accessible and cost-effective compared to traditional laser scanning. Archaeologists use IBM for site recording, monitoring changes over time, and creating virtual reconstructions, which aid both research and public engagement. As the technology advances, IBM is increasingly being combined with AI-driven interpretation and augmented reality (AR) applications to enhance both academic study, museum experiences, and online dissemination [27,28]. The typical IBM procedure involves several steps: alignment, 3D point cloud generation via SfM photogrammetry, 3D mesh reconstruction, and texture projection, which involves projecting colours from the images onto the 3D mesh.

For this project, 4056 × 3040 pixel RBG images were captured using a DJI Flir2 HT camera mounted beneath a DJI Matrice 300 UAV. The 3D models were created using RealityCapture (v.1.2) software by Capturing Reality, Bratislava, Slovakia. Simplified versions of these models were then analysed using the ASCII technique in the GigaMesh software (v.240221), developed by Austrian computer scientist Hubert Mara [29]. This analysis and further manipulation produced height maps highlighting terrain potentially containing archaeological features.

2.3. Thermal Photography

Simultaneously, as the aforementioned RGB images were taken, the UAV-mounted camera photographed thermal images (Figure 3A). These images have a resolution of only 650 × 512 pixels, requiring flights to be conducted at a low altitude with a high degree of overlap. Thermal imagery, or thermography, captures the infrared radiation emitted by surfaces and converts it into a visual representation called a thermogram. The technique originated in the 1920s, but the production of real-time thermal imaging systems was made possible by advancements in detector technology and signal processing in the mid-20th century [30] (pp. 1–24). Because of its non-destructive potential to discover hidden archaeological features by detecting variations in surface temperatures, thermography has emerged as a valuable tool for archaeology. For instance, the technique has been used for site prospection, mapping of ancient structures, and monitoring of environmental changes affecting archaeological sites [31,32].

For this project, several steps were undertaken to generate large-scale thermal images for potentially detecting significant anomalies at the sites. Initially, the original images needed conversion from a format usable only with DJIs proprietary software to a more versatile format. This was accomplished using ThermoConverter (v.1.5) software developed by Vantage UAV Limited, Bristol, UK. The resultant JPG images were then imported into Thermal Studio by FLIR. Within this software, entire batches of thermal images could be manipulated to highlight only the optimal temperature ranges for each site. These altered images were then exported using various colour palettes to examine which best showcased potential archaeological features. Thermal Studio cannot produce panoramas comprising hundreds of images, so 3D models were generated based on greyscale thermal images using the RealityCapture software. These 3D models could subsequently be overlaid with different temperature scales and colour palettes using the batches of altered images exported from Thermal Studio. The resulting different 3D models and orthophotos could then be manually inspected for potential archaeological anomalies.

2.4. Topsoil Magnetic Susceptibility (MS) Survey

Magnetic susceptibility (MS) is influenced by the presence of magnetic minerals in the soil and a property obtained by generating a low magnetic field and measuring how much more magnetic a material becomes under the influence of this field. Measurements can be conducted directly in the field, as topsoil measurements with a loop or downhole sampling, or as laboratory analysis of soil samples taken from sections or coring [33,34]. Not all types of soil have the same content of these minerals, and they can be altered by activities such as burning, biological decay, bioturbation, and temperature fluctuations. Since human settlements often involve several of these processes, systematic measurements can be used to locate and delineate zones of human activity. Studies in Norway have particularly focused on sites related to iron production [35,36,37] and iron-age settlements [38,39,40]. The introduction of anthropogenic materials like iron slag, ceramics or industrial waste can further increase the availability of magnetic minerals at archaeological sites and within archaeological features [2,33,34]. However, measurements can also reflect natural variations in the subsurface, especially where there is significant soil variation. Therefore, interpreting the results must always take local conditions into account.

For this investigation, a Bartington MS2 with a D-loop was employed. The sensor measures volume susceptibility up to 10–15 cm below the coil, with a primary response deriving from the upper 6 cm of this sample volume [2,3,41]. Recorded values are relative (SI) and documented on a 1.0 scale. This provides a quick way to measure susceptibility in the field [33] (p. 10). Measurements were taken on a regular 5 m × 5 m grid to compromise between the necessary resolution and efficiency and to comply with the half-day time limit. This resolution has proven sufficient to delineate iron production and settlement sites [36,38]. Interpolation was carried out by kriging in ArcGIS Pro, with the total number of the measures, the interquartile range (IQR), and 2 standard deviations (2SD) of the mean.

2.5. Fluxgate Gradiometer Survey

The contrast in magnetic susceptibility between archaeological features and the surrounding mass determines whether these can be detected by using a magnetometer. Further, by using two magnetometers mounted above each other, one can eliminate the effect of the Earth’s magnetic field, allowing measurement of variations in the strength of a local magnetic field caused by the magnetisation of features beneath the ground. This mode of magnetic surveying is known as gradiometry (Figure 3B).

The method is considered suitable for detecting traces of burning or industrial activity, but the range of features detected includes ditches, pits, hearths, or stone features. It can achieve relatively high spatial resolution, potentially identifying features with diameters as small as 0.5 m depending on crossline spacing and configuration.

In northern regions, a structure filled with more magnetically susceptible material than its surroundings will yield measurements with positive values slightly shifted toward the south relative to the centre of the structure and a negative part of the signal towards the magnetic north. This is generally referred to as induced magnetic behaviour. For self-magnetised or magnetically remanent materials, however, the negative measurement can occur in any direction. Materials containing iron minerals, for example, can exhibit remanent magnetism when exposed to temperatures above approximately 550–600 °C (the Curie temperature) and then cooled down. This phenomenon is known as thermoremanent magnetism. Archaeological features such as burnt stones, hearths, iron objects, or slag may exhibit positive and negative measurements that do not necessarily align with the N–S direction and where the negative part of the signal can be significant [42,43].

The fluxgate gradiometer surveys were conducted using a Bartington Grad 601 system with two fluxgate gradiometer sensors. The data were collected in zigzag mode, with adjacent survey lines being in the opposite directions. The line spacing was 0.5 m while the station spacing was 0.125 m. The data were processed in Terra Surveyor by using Destripe (Median Traverse, threshold: 2.5 SD), Despike (3 intervals, 1 threshold, mean), Low pass Gaussian filter (3 × 3), and Interpolation. The grids were later georeferenced in ArcGIS Pro, according to precise CPOS GNSS measures of the grids.

2.6. Ground Penetrating Radar (GPR)

Detailed images of the subsurface can be created by sending electromagnetic (EM) energy into the ground and measuring the time it takes for some of that energy to be reflected to a receiver. These reflections are primarily caused by changes in the material’s dielectric permittivity, i.e., a material's ability to store a charge from an applied electromagnetic field. These conditions are closely related to conductivity influenced by moisture content, with a minor contribution from differences in magnetic properties [44]. Typically, very moist subsurface conditions attenuate more of the signal, resulting in lower geophysical contrasts, whereas stone-filled structures in otherwise homogeneous sediments provide excellent geophysical contrasts and are relatively easy to detect [2,44,45]. This method is thus well-suited for detecting stone structures, ditches, pits, and postholes, but the detected features should still be considered a minimum of what is visible in the data and present in the ground [2,3,38]. At complex sites such as the shielings, mainly consisting of thin deposits in rocky moraine substrata—and assumably with a high degree of reuse—it may be more problematic to discern archaeological features.

The grids were positioned in central areas, considered to have the highest archaeological potential, yet site topography and access also affected the location. The project employed a Malå GX antenna with a centre frequency of 450 MHz, generally considered well-suited for detecting archaeological features (Figure 3C). Data were collected in a zigzag manner where adjacent survey lines, separated by 0.25 m, pointed in the opposite directions. Conducting GPR surveys on a series of parallel survey lines allows the creation of depth-specific maps, called depth slices, which may reveal the extent of subsurface features like house foundations and stone walls. The two-way travel time of the EM wave is measured in nanoseconds. By analysing specific signal characteristics, approximate depth estimates in centimetres can be made based on the velocity of the signals in different materials. However, it is essential to recognise that these depths are estimates, as EM wave velocities can vary depending on material and water content [44] (pp. 56–58).

The collected datasets were processed in GPR Slice using T0-correction, bandpass filtering, background removal, and migration, in that order. The interpolation was carried out by kriging, and the program also allows for direct profile and slice correlations in 3D. The grids were georeferenced in ArcGIS Pro based on precise CPOS GNSS measures from every corner and from orthophotos created using UAVs.

2.7. Soil Coring

Based on the results from the non-invasive investigations, grids were laid out across key areas, and systematic soil coring was carried out every 2 m × 2 m. For ground truthing purposes, the aim was to cover a representative sample of anomalies from the different remote sensing methods.

As a method, soil coring is fairly uncommon in Norwegian archaeology, as in many other parts of Europe, but could be regarded as a less invasive technique with minimal site disturbance. By capturing perceptible soil-colour variations, archaeological deposits can be differentiated from modern or natural deposits [46]. Recent advances include the use of interpolation techniques and sub-surface topographic modelling [47].

The investigations were conducted with a simple, lightweight soil probe. The depth to subsoil, typically between 30 and 50 cm, was measured and manually noted at each point, along with the thickness of charcoal layers or distinct archaeological deposits. Using interpolation (kriging) in ArcGIS Pro, depth maps of the topsoil were created for each site, along with the spatial extent of any clearly discernible past horizons.

2.8. Manual Trial Trenching

The excavation permit allowed a maximum area of 6 m² to be opened at each site. Typically, two 1 m × 3 m trial trenches were located across potential buildings or other archaeological features. The positions and orientations of the trenches were based on a combined assessment of visual surveys, non-invasive investigations, and subsequent soil coring and aimed at features detected by various remote sensing methods.

One half of each trench was stratigraphically excavated, reserving the other half for sampling. In cases of thicker strata, the soil was uncovered in 5 cm slices. Small stones were removed, while larger stones—potentially part of walls or other anthropogenic features—were left in place. Artefacts were collected, but no sieving was carried out due to low expectations. Because the utensils used in the shielings were mainly wooden, the sites rarely contain any preserved artefacts [7,48].

Additional test pits or small trenches were prioritised at some sites to clarify possible features and contribute to better site delineation, rather than excavating the full extent of the 1 m × 3 m trenches. When finally excavated, the trenches were documented by IBM. The soil and turf were carefully filled back into the trenches, and great care was taken to restore the landscape to its former state after the investigation.

3. Results

The main ambition in the project has been to assess the potential for various remote sensing and non-invasive methods in pastoral mountain landscapes, both in terms of archaeological results and practical applicability within cultural heritage management and developer-led archaeology. Yet, the shielings are not a homogenous group of sites when it comes to local ground conditions.

A measure that may capture varying terrain surface properties, i.e., whether the ground was smooth or rough—consisting of homogenous grass and turf or a heterogenous mixture that includes stones, tussocks, and boggy spots—is the variation or span of the topsoil magnetic susceptibility measures (SI) per m². Before excavating the 1 m × 3 m trial trenches, we ran a test by measuring MS every 0.25 m × 0.25 m within this smaller grid. Generally, the mean span per m² and the number of outliers align with the subjective experience of the sites and their physical appearance. The range of these readings also impacts the uncertainty or randomness of the more extensive topsoil mapping conducted across the sites at lower resolution (5 m × 5 m). Søstølen may be an exception due to the extremely high MS values, which further influence the span (

Table 2. Mean span of Volume Susceptibility (SI) per m² in grids across the planned trial trenches.

Site (Trench)	Mean SI Span (IQR)	Mean SI Span (SD)	Outliers (n)
Klovset (T3)	2.00	6.67	1
Klovset (T2)	4.00	9.33	2
Setersetran (T1)	5.00	12.67	2
Setersetran (T0)	7.17	16.67	4
Myrset (T1)	8.67	5.67	1
Myrset (T2)	4.67	13.00	7
Myrset (T0)	12.00	20.00	4
Rangsetra (T1)	11.50	26.00	4
Søstølen (T2)	22.33	56.67	0
Søstølen (T1)	62.25	180.00	3

, Figure 4).

Naturally, operating aerial surveys is less dependent on terrain properties. As none of the sites were densely forested, the practical conditions and applicability were fairly similar across all five sites. LiDAR data from the sites was collected in a few minutes but extended to cover more significant portions of the surrounding landscape. The use of the UAV with cameras was somewhat more time-consuming, although it still proved that entire shielings could be surveyed within half a day. However, UAVs depend on battery capacity, favourable weather conditions, and, for the thermal camera, even on solar conditions and the time of day. The conditions were generally good for geophysical prospection, with mostly dry weather during and prior to the investigations.

As for the LiDAR, there are some cases of striping artefacts occurring in the datasets. The artefacts result from sudden stops in the forward movement of the drone, caused by height adjustments in the terrain-following software, during data collection. The stops cause the drone and sensor to wobble back and forth, producing duplicated data with small height misalignments. This was caused by the settings of the flight program, where the height adjustment operation was set to a “stop and turn” movement, whereas it should have been a “gradual” movement, eliminating the wobble of the sensor. This was not realised until after the post-processing of the data. The problem is most visible in areas with significant changes in topography, such as at Rangsetra, where we opted to process only half of the flight grid to avoid the flight lines with the most topographical variations.

In the following subchapters, the results of the different survey methods are presented, shieling by shieling, starting with the mildest terrain and the most well-maintained site. All the sites have followed the same workflow (Figure 5).

3.1. Klovset

Klovset is a historically known shieling, situated at c. 550 m above sea level (m.a.s.l.) in the mountains near Valldal in the inner parts of the county (7°12,86′E, 62°17,77′N), and associated with the farms by the fjord. The suffix of the site name (Old Norse setr) suggests a Late Iron Age date and the possibility that the site had permanent settlement [49] (pp. 136–145). The prefix (‘cloven hooves’) may either refer to the significance of livestock or pack animals or to its location between two rocky knolls, only covered by thin moraine deposits. Most of the current buildings are fairly recent, as an 1862 century allotment map displays their former location, arranged in a row along the front ridge (a–e, Figure 6). The ruins of barns and byres are spread across the site (f–k), some of which are also marked on the mid-19th century map. The NW part of the site has been subjected to modern intervention, and building k is no longer visible. The enclosed meadows were mown, and the hay was transported down to the farm by a rope and pulley system. The site is still open, green, and lush, as there are sheep grazing in the area, and the meadows receive annual upkeep to enhance the recreational value of the site. In 2016, the first archaeological investigations were conducted at Klovset, in connection with a planned cabin on the SW part of the site. Inside the enclosure, test pits revealed archaeological deposits starting from the Merovingian period (AD 584–658, 1430 ± 30 BP, Beta 439736). Hence, the date corresponds well with the setr-name. As the investigation was carried out at the edge of the site, the central area was expected to have a potential for more substantial archaeological features [50] (p. 13).

3.1.1. Aerial Survey

Both LiDAR and IBM displayed most of the 19th-century house foundations (a–j), and in addition, two similar depressions which could not be accounted for in the historical maps (m–n). Both were located between the two ridges and in the slope facing NE. They are both slightly visible in the terrain but were not accounted for prior to the aerial surveys. In addition, the manipulated local height models also displayed a larger building (o) in the W part of the site.

The UAV-based surveys at Klovset were carried out by night after a long, sunny day, providing good conditions for the thermal camera. High-temperature variations were captured, from wet grasslands to sun-warmed stones and rock outcrops, and features such as the enclosure (l) were clearly visible. However, the remnants of the dwelling houses from the 1862 map (a–e), all covered with soil and grass, were not detected by thermography. The only subsurface house structure detected by this method (p) was located on the NE slope, correlating with an elevated terrace that was visible in the LiDAR and IBM data and tentatively interpreted as an 18th or 19th century byre.

3.1.2. Geophysical Prospection

The conditions for geophysical surveys were generally good (Figure 7), with relatively flat and mild terrain, low vegetation, and few obstacles (except for present houses and rock outcrops).

Measuring topsoil magnetic susceptibility, an area of 8600 m² was covered during half a day, as the work was often disrupted by interested hikers and tourists that came to see the fjord view. When presented as 2SD of the mean, the SI values display some spikes around the present buildings. There are also elevated values along the ridge with the 19th-century house foundations (a–e), some of which could be affected either by sub-terranean rock formations or modern camping. Conversely, the lower values measured in between the ridges may be caused by deeper deposits. Early settlement horizons may thus be covered with soil and beyond the MS2D depth range.

The gradiometer survey was conducted in three 30 m × 30 m grids in open areas between the buildings and projecting rock outcrops (Figure 8). The rock formations dominated the dataset, exhibiting strong dipolar readings with an irregular shape. However, two dipolar features (q–r) spatially related to the two depressions visible in the IBM and LiDAR datasets (m–n) are interpreted as anthropogenic features. There may also be other features (s), but they are hard to discern from erratic magnetic rocks close to the surface.

A GPR grid of 40 m × 27 m (1080 m²) was collected in three hours. The grid was placed in the open area in front of the standing buildings (Figure 9), covering the 19th-century house foundations (a–e). Despite their known location, these were surprisingly difficult to recognise, perhaps because of the undulating terrain. As suggested above, the GPR profiles measured deeper deposits between the two rock outcrops. Some apparent features were detected in the depth slices, like a 2 m-wide dense concentration of strong point reflections (t), interpreted as a large cooking pit (Figure 9), and something that looked like the corner of a house foundation in the N corner of the grid (u). Neither was clearly discerned in the gradiometer data. In addition, there were several possible posthole pits and house foundations (v–w), but the moraine subsoil made them hard to distinguish from naturally occurring stone.

3.1.3. Soil Coring and Trial Trenching

The coring at Klovset covered an area of 880 m², encompassing most of the features identified by aerial survey and geophysical prospection (Figure 10). The results showed a similar pattern to the GPR survey, with topsoil depths over 30 cm in the rift between the two rocky knolls. In addition, areas with more distinct charcoal-rich deposits were detected, particularly in the NE slope. By anticipating that these deposits are ancient, it is possible to discern early activities from more recent soil accumulation. The activity area was not detected by magnetic susceptibility measures due to the limited depth range.

In addition to opening a small trench to verify the possible cooking pit (t) revealed by the GPR prospection, trial trenches were placed across two possible house foundations (u and m). Both were placed within the areas with identified archaeological deposits. The first trench (T1) soon reached a compact layer of fire-cracked stone. It did not contain charcoal, and rather than a cooking pit, it appeared as merely a pit of loose, fire-cracked stones (Figure 11). The second trench (T2) was placed across the wall of a possible house foundation (u), identified by GPR as two perpendicular walls in the corner of the grid. Both walls were recovered in the 0.5 m × 3 m trench, as one of them happened to follow the southern section. The opposite section displayed stratified deposits, and a sample from a charcoal layer that went underneath the wall was dated to AD 677–818 (1260 ± 13 BP, TRa-21646). Below, there was an even older archaeological deposit, containing a lump of iron slag.

The last trench (T3) was located across the outer limit of one of the two visual depressions (m) in the charcoal-rich area facing NE. The trench exhibited a stone row where the wall was expected to be (Figure 12) and a supposed occupation or floor level which was dated to AD 1046–1201 (914 ± 12 BP, TRa-21647). Beneath the floor, however, there was a distinct pit formation filled with charcoal and rocks, including larger stones that were not fire cracked. This was only slightly older, AD 1023–1151 (979 ± 14 BP, TRa-23823), and is probably contemporary with the building. The interpretation is unclear, but it may represent a fireplace or some kind of production related to the two dipolar magnetic features in the gradiometer data and the pit in trench 1. Altogether, the investigations at Klovset proved very conclusive, and most of the non-invasive methods yielded viable results.

3.2. Setersetran

Setersetran is located in a side valley from Øksendal in the northern part of the county, at c. 435 m.a.s.l. and in a downward slope facing NW (8°26,71′E, 62°40,04′N). The site is enclosed and features two standing timber dwellings, two byres and a deteriorated 19th-century hay barn (a, Figure 13). The enclosed meadow (b) remains open, adorned with grass and blueberry shrubs. Geologically, the subsoil consists of glacio-fluvial deposits. In 2021, a short archaeological survey was conducted after the landowner had found a whetstone inside what was interpreted as a house foundation (c). Ceramic fragments suggested a recent date, but farther downslope, approximately 30 m NW of the standing buildings, there was an overgrown stone wall (d) that indicated an earlier phase. The meadow’s flattest, smoothest, and most fertile section is located below this wall, and patches of a charcoal layer were found here and dated to the transition between the Migration and Merovingian periods (AD 485–602, 1521 ± 19 BP, TRa 16834). Some terraces (e–f) close to the wall may suggest settlement, but no definite Iron Age or Mediaeval features were known prior to our investigations. Features of varying certainty were also noted close to the current buildings at the top of the hill (g–j), but no archaeological deposits were found. Hence, the possible house foundations were either considered modern or natural.

3.2.1. Aerial Survey

The UAV surveys at Setersetran were affected by dense shrub vegetation, blurring the microtopography and ground temperature measurements. Further, the investigations were conducted in the afternoon, with the sun still hanging low. This resulted in relatively poor conditions for thermography, as the terrain was heavily affected by sunlight, low vegetation and shadow effects.

The LiDAR and IBM surveys were more credible, although the blueberry shrubs hampered the resolution across most of the site. In addition to the stone wall and other clearly visible features, some terraces and pit formations were interpreted as possible subterranean house foundations. This included features interpreted as possible houses during the initial survey in 2021 (e–f) and some depressions on the grassy ridge below (k–l).

3.2.2. Geophysical Prospection

The geophysical prospection at Setersetran (Figure 14) was conducted under varying weather conditions. The magnetic surveys were carried out during light rain, whereas long grass in the autumn hampered the GPR survey.

The magnetic susceptibility survey was effective and covered an area of 11, 175 m². Areas surrounding the present buildings were clearly marked as hotspots, but elevated measures along the trail could suggest a wider activity area. Whereas the slope below the buildings exhibited low SI values, there were elevations across the smoother surface below the ancient stone wall. Based on the 5 m × 5 m resolution, however, it is not feasible to detect specific features.

The gradiometer survey was conducted in four grids of 30 m × 30 m (3600 m²), covering central areas below the stone wall and where the farmer had located a house foundation (c) in 2020. The stone wall is visible in the processed dataset, but the possible dwellings (c, e, f, j, l) are more challenging to discern. The magnetic response is highly affected by geological formations—visible as amorphous zones of strong positive and negative responses. One of these zones, however, was located right below the possible building (e) along the wall and may have been formed by human activity.

Due to terrain constraints, with slopes, tussocks, and shrubs, only a small GPR grid was collected. Measuring 20 m × 20 m (400 m²), it was placed across the grassy ridge mentioned above. The possible house foundations discovered by the aerial surveys (k–l) were also visible in the GPR data, measuring c. 4.5 m × 4.5 m.

3.2.3. Soil Coring and Trial Trenching

At Setersetran, the soil coring covered an area of 960 m², including the ancient stone wall and the fertile meadows below (Figure 15). The upper part of the grid, generally consisting of rough, sloping terrain, generally displayed low soil depths. The exception was right below the house foundation (e) by the stone wall, correlating with the magnetic response from the gradiometer survey (Figure 16). Further downhill, the meadow generally exhibited thicker deposits, with some charcoal patches and archaeological contexts. Somewhat surprisingly, however, there was only a thin layer of topsoil across the aforementioned ridge (k, l). More substantial archaeological deposits were found 15 m further S, just below the slope. Upon closer inspection, an elongated depression was found above these deposits, interpreted as a possible house foundation (m). This is visible in the LiDAR and IBM data but was initially interpreted as a natural depression. A test pit confirmed distinct archaeological deposits.

Initially, the trenches at Setersetran primarily aimed at two features: the terrace or plateau with the Iron Age date (f), further highlighted by LiDAR and IBM, and the upper of the two depressions (k) on the grassy ridge below. The latter was detected by several methods. As the coring along the ridge did not reveal any archaeological deposits, two test pits were opened at both ends of the planned trench to assess the further potential. As none contained archaeological horizons, this trench was dismissed.

The first trench (T1) soon revealed the charcoal layer from the late Migration/early Merovingian period, but there was no evidence of constructions nor any substantial archaeological deposits. The platform appeared natural, and the charcoal may be related to early fire clearance. The second trench (T2) was relocated to the newly discovered house foundation (m). In addition to the archaeological deposits, a stone wall construction was located along the rim of the depression. However, the deposit turned out to be post-mediaeval (83 ± 14 BP, Tra-21642).

The overall interpretation is that there was early activity in the area by the end of the Early Iron Age, with land clearing and possible transhumant activity. However, the house foundations and the stone wall are more likely related to early modern summer farming. The dated building (m) appears to be a dwelling, whereas the building along the old enclosure (e) may have served as a cattle byre. The deposits contained significantly less charcoal, and the deep soils below may have been formed by past dung heaps. The location along the stone wall is quite typical [7] (pp. 345–346), and the cattle byres still standing have a similar relation to the more recent enclosure. As the surrounding birch forest was cleared, the enclosure may have thus been extended and the houses moved uphill.

3.3. Myrset

Myrset is located in the central part of the county and situated below a hilltop at c. 365 m.a.s.l. (7°33,28′E, 62°39,31′N). The prefix Myr- refers to the surrounding bogs, but the site itself is situated on a small, thin pocket of glacial moraine deposits. Similar to Klovset, the suffix suggests a possible Late Iron Age farm settlement [49] (p. 144), and there is also a local tradition of people having lived at the site permanently. All the standing houses are recent and date from the 20th century, but older house foundations of varying certainty are still visible (a–l, Figure 17). Surrounding the site is a distinct enclosure (m). There is still a grassy clearing in the central parts of the site, but juniper and deciduous woodland vegetation is slowly encroaching. Myrset has been investigated several times, with numerous radiocarbon dates recorded from various contexts [51,52,53]. It is interpreted to have been a marginal farmstead during the Viking Age and Early Middle Ages, with traces of bog iron production and other activities going back to the Early Iron Age. Pollen analysis has also revealed traces of barley [7] (p. 350). The possible floor of a subterranean house foundation (n) from the 12th century (AD 1042–1219, 900 ± 30 BP, Beta 323546) was discovered by test-pitting during 2014. Most likely, the building was related to nearby iron production. By the 13th century, however, the production seems to have ceased, and Myrset later turned into a seasonal shieling [52,53].

3.3.1. Aerial Survey

Despite poor conditions with strong winds and sunshine obstructing the thermal investigation, most foundations are visible in the datasets from the UAV-based survey methods. The same goes for the stone wall surrounding the site (m). Finding prehistoric or mediaeval house foundations, however, is far more challenging. The thermal camera picked up some rows of stones south of the recent ruins, interpreted as a possible building (o), but no new features were detected with certainty.

3.3.2. Geophysical Prospection

The magnetic surveys at Myrset were interrupted by a thunderstorm and had to be completed another day. Within a total of approximately four hours, an area of 14,200 m² was covered by magnetic susceptibility measures. By investigating an extensive area, we hoped to locate any traces of metalworking in the area surrounding the main settlement. The SI values are generally low due to the boggy surroundings, with some very high and deviating values (150–350) close to the present buildings. By only visualising the data range of 2SD of the mean, we could locate a belt of high values surrounding the site and the central settlement area. Modern cabins in this area may account for some of the highest values. However, the continuous belt that marks the boundary between the site and the surrounding bogs may suggest prehistoric activity related to iron production (Figure 18).

During the gradiometer survey, three 30 m × 30 m grids (2700 m²) were collected in the open areas surrounding the main settlement area. Some of the barely visible house foundations, including the one identified by thermography (o), are slightly visible as anomalies with multiple small positive and negative readings in the SW grid. Besides, there are stronger dipolar responses related to one of the recent ruins (f) and just below the buildings (p), but nothing further noticed in the data indicates any large-scale iron production within the collected grids. Some strong bands, most probably derived from geological formations, are going through the dataset.

Due to rough terrain, the GPR survey only covered a smaller area measuring 42 m × 15 m (630 m²). In the upper depth slices, at about 0–25 cm, two distinct buildings were detected, one of which was surveyed in 2004 (l) and dated to early modern times [51] (p. 47). The second (q)—located next to one of the magnetic anomalies mentioned above (p)—was entirely overgrown but orientated in the same direction. The GPR data showed that both had a narrow depression orientated along the centre and are interpreted as post-mediaeval cattle byres. Lower slices were more challenging to interpret but included an irregular linear feature (r) below the two byres that seemed to be of archaeological interest.

3.3.3. Soil Coring and Trial Trenching

Systematic coring at Myrset covered an area of 280 m², emphasising the area just below the main historical settlement area. This included one of the cattle byres (l) and the linear feature (r) uncovered by GPR (Figure 19). The 2004 survey exposed substantial archaeological deposits underneath the floor of the byre [51] (p. 47), and the soil coring helped delimit these deposits. Their edge clearly correlated with the linear feature detected by GPR.

A trench (T1) was opened across this linear feature (r), assuming it could be an enclosure or a demarcation of some kind. Underneath a more organic layer, distinct archaeological deposits were encountered in the upper part of the trench. Whereas these horizons were stone-free, a pile of stone was found by the lower end—interpreted as the result of clearance. Hence, the deposits above are interpreted as remnants of cultivated fields, suggesting permanent farm settlement. Similar deposits were encountered during the 2013–2015 investigations and dated to the Viking Age and Early Middle Ages [52,53]. Underneath these horizons was a pit containing charcoal and iron slag, further demonstrating metalworking activities prior to settlement. A new sample from the organic layer on top suggests the site was abandoned in the 14th century, around the time of the Black Death (662 ± 14 BP, TRa-23822).

The second trench (T2) was initially planned across the feature (o) revealed by the thermal camera and gradiometer, but soil coring indicated that the feature was primarily composed of stones. Furthermore, due to the location—outside the area with archaeological deposits and right next to recent house foundations—it was rather decided to place the second trench across the possible building (n) identified by soil coring and test pitting in 2013 [52]. Only a shallow depression is visible on the surface, but a line of stones was soon detected along one of the proposed wall ridges. Inside, the Early Mediaeval floor deposit was rediscovered. It contained some smaller artefacts, such as whetstones and fire flint, and was clearly delimited by the wall structure.

3.4. Søstølen

Søstølen is an abandoned farm or shieling in the Eidsdal valley (7°9,81′E, 62°10,51′N), close to the UNESCO World Heritage Landscape of the Geirangerfjord in the southern and innermost part of the county. The site is located on a hill at c. 495 m.a.s.l. and about 600 m from the uppermost farm in the valley. Thus, permanent settlement at the site cannot be ruled out. The place name (from Old Norse stoðull, ‘milking site’), suggesting transhumant activity, is not directly linked to the site but refers to a rock formation in the mountain above. The local name of the site is Haugane, ‘the hills’. The area is still intensively grazed by cows, sheep, and goats and has a lush and verdant appearance, with no forest cover (Figure 20). Geologically, the area is covered by thick glacial moraine, but numerous large boulders bear witness to a significant rockfall risk.

The site consists of three horseshoe-shaped house foundations (a–c)—one of which has been hit by one of the boulders—and a possible stone wall (d). Archaeologists first visited the site in 2014 as part of a developer-led survey related to road construction and avalanche protection measures [54]. A charcoal sample was obtained from one of the ruins (b) and dated to the Viking Age (AD 882–1015, 1110 ± 30 BP, Beta 394048).

3.4.1. Aerial Survey

The house foundations (a–c) are clearly exhibited in the LiDAR and IBM datasets. The UAV based surveys at Søstølen were conducted in the afternoon, after the sun set behind the steep mountains, which provided good conditions for thermal photography. In addition to the house foundations, the thermograms also highlight the possible stone wall (d) and the boulders from continuous rockfalls. Combining LiDAR and IBM with thermography provides a good overview of the visible features, but—except for a seemingly modern fireplace (e) in the turf layer—no subsurface features were detected.

3.4.2. Geophysical Prospection

Within two hours, measures of magnetic susceptibility covered an area of 8900 m². Good conditions would have allowed for a larger coverage, but the grid encompassed all known features and was considered sufficient (Figure 21). The SI values were quite high across the site, with an interquartile range between 22 and 69. A boggy area to the south of the house foundations had values around 0, whereas the highest levels were found around and in between the house foundations.

The gradiometer survey effectively covered four grids in a 60 m × 60 m (3600 m²) square in the open terrain, covering all three house foundations. Due to the challenging geology, however, there is much noise. The strongest response derives from a stone or rockfall that is visible on the surface. Hence, anthropogenic features are difficult to isolate. The stone wall is discernible, but even visible house foundations were hard to recognise.

The GPR survey was challenging due to the rough surface. The 20 m × 30 m grid (600 m²) covered one of the house foundations and the area where the visible stone wall was expected to continue. The house foundation was clearly visible, as well as a scatter of stones (f) in front of the building—perhaps the remnants of a collapsed stone-walled gable. Some possible features were identified at deeper levels, notably a linear reflection (g) running perpendicular to the visible stone wall. This was interpreted as a possible enclosure.

3.4.3. Soil Coring and Trial Trenching

Soil coring every 2 m × 2 m in a 17 m × 20 m grid (340 m²) was undertaken in the same area as the GPR prospection (Figure 22). Substantial deposits were concentrated in the inside and front of the building (a). Surrounding areas had only 10–15 cm of soil coverage and were even thinner below the possible stone wall (g). However, one area between the three ruins, with deeper deposits and distinct charcoal horizons, may reflect possible subterranean archaeological features.

The trial trenches focused on one of the house ruins (a), as well as the possible stone enclosure (g) that was detected by GPR. The evidence for the building was solid, clearly visible, and recognisable by most methods. A trench (T1) was laid across the possible opening in the house wall, mainly for dating and environmental sampling. Archaeological deposits were soon detected both inside and in front of the ruin, as well as the foundations of a gable wall and numerous stones scattered outside (f), as suggested by the GPR prospection (Figure 23). One of the floor deposits inside the house was dated to AD 1177–1262 (835 ± 13 BP, TRa-21645). The trench across the possible enclosure (T2) was negative, and the weak linear reflection (g) appeared geological. Later, we were told by a local farmer that the visible stone wall was the result of modern stone clearing.

3.5. Rangsetra

Rangsetra is another abandoned shieling. It is situated at c. 325 m.a.s.l. on a slope facing E in the Høydalen valley in the Austefjord area of southern Sunnmøre (6°11,59′E, 62°4,16′N). Presently, no buildings occupy the site, but the name suggests summer farming. Since the early 17th century, two small farms have been situated just below the site [55] (pp. 157–163). According to local traditions, Rangsetra was used by these farms, most likely during springtime [56] (p. 104). However, uncertainty surrounds the dating of the two nearby farms and whether Rangsetra rather belonged to older farms by the fjord. Situated on thin moraine deposits, strewn with rocks and hillocks, the site has an uneven appearance and a rough terrain that is not known to have been mown. As the area still serves as grazing land for sheep, Rangsetra has largely remained devoid of forests, with only scattered shrubs and small trees, including some recently self-sown spruce trees from an old plantation nearby. The area was archaeologically surveyed in 2009, during fieldwork related to the 420 kV power line between Ørskog and Fardal. Possible house foundations were located but poorly mapped (a, Figure 24). Radiocarbon dates from charcoal layers nearby were dated to the Early and High Middle Ages (880 ± 30, Tra-32, 1000 ± 30, Tra-33).

3.5.1. Aerial Survey

The aerial surveys had some difficulty detecting any of the barely visible features (a–b) at Rangsetra. As the survey was conducted during the middle of a sunny day, the conditions were poor for thermography. Based on processed LiDAR data, two possible house foundations were outlined (a–b), as well as an already known stone cairn (c). This was mainly because the same features were located on the ground and on the basis of photos from the initial survey report [56]. Based solely on the DEMs, it would have been impossible to obtain convincing results in the very rough terrain. According to tradition, there was also supposed to be a tar kiln in the area. These are typically distinct features in LiDAR and IBM datasets, yet none were spotted in the first round of surveys (Figure 24).

3.5.2. Geophysical Prospection

Magnetic susceptibility was measured over an area of 2300 m². The general conditions were good and would have allowed for a somewhat larger coverage, but the main area was covered. However, the rough terrain with rocks and tussocks yielded some deviating high results (140–150), most likely erratic magnetic rocks close to the surface. By removing the spikes (2SD), the observed pattern turned out opposite of what to expect. The area supposed to contain archaeological features (a, b) had very low values. In contrast, the highest values were obtained in the SW outskirts of the grid. This could be due to modern activity in the area, but the nearby 420 kV powerline may also have influenced the values.

The gradiometer survey also faced problems in this rough terrain. Firstly, it was difficult to keep a steady pace during data collection, and only one 30m × 30 m grid (900 m²) was managed within the designated time frame (Figure 25). This covered the supposed settlement area, but not the tar kiln (as it was not yet discovered). Secondly, the acquired dataset was quite noisy and yielded no significant and clearly identifiable results.

Due to the rough terrain, GPR prospection was not feasible at Rangsetra. Hence, none of the geophysical methods provided any viable archaeological results, and the site would not have been detected without visual inspections and soil coring.

3.5.3. Soil Coring and Trial Trenching

The visual survey identified two possible house foundations in the supposed settlement area. A shallow depression at the top of the hill (b) seemed more promising than the one surveyed and photographed in 2009 (a). As no deposits were found inside this building during the initial survey, the charcoal samples were retrieved somewhat further north.

Coring was conducted for every 2 × 2 m in a 26 m × 16 m grid (416 m²), covering the plateau and the two possible house foundations (Figure 26). The survey confirmed the lack of deposits within the lower building (a). Interestingly, however, the northernmost part of the grid contained quite dense charcoal deposits. This led to the discovery of an overgrown tar kiln (d) just above the grid, and a suspicion that the radiocarbon dates from 2009, although from birch, could have been related to this activity. However, the second house foundation (b) proved more convincing, with substantial archaeological deposits inside and in front of the visible depression. Moreover, the charcoal content was considerably lower than by the tar kiln.

The trenches were initially planned across the two possible house foundations, but due to the results from the soil coring, the southernmost (a) was omitted, and a test pit was rather dug into the tar kiln (d). To avoid contamination from tar production, the trench across the other house foundation was moved to the S wall.

The trench (T1) soon unravelled a solid stone wall, as well as distinct archaeological deposits inside the depression (Figure 27). They were, however, covered with very organic deposits, which probably explained the negative results from the magnetic surveys. Furthermore, a 25 cm deep posthole pit was found in the inner end of the trench, suggesting that the building was an earth-dug post construction with solid stone walls and the gable facing downhill. The posthole pit was dated to AD 1229–1280 (764 ± 13, Tra-21643).

4. Discussion

From the results above, encompassing a broad range of sites and landscapes, we can discuss the potential and practical applicability of remote sensing methods at the shielings and within the context of cultural heritage management and developer-led archaeology.

4.1. Practical Applicability—Access and Efficiency

The results above clearly show the efficiency of aerial surveys, particularly LiDAR (

Table 3. Results—efficiency and coverage (measured in hectars) with various methods, within half-a-day time frame.

Site	LiDAR	Thermal	IBM	MS	Gradiom.	GPR	Coring
Klovset	40 *	3.20	3.20	0.86	0.27	0.11	0.09
Setersetran	30 *	1.51	1.51	1.12	0.36	0.04 *	0.10
Myrset	50 *	1.62	1.62	1.42	0.27	0.06	0.04
Søstølen	30 *	3.05	3.05	0.89	0.36	0.06	0.06
Rangsetra	30 *	1.58	1.58	0.23 *	0.09	-	0.04

* The methods could have covered larger areas, but the was extent was considered sufficient for the problems raised.

). Furthermore, as the UAVs are airborne, they can travel far and access remote areas where open access ALS coverage is poor. Operating within a VLOS (Visual Line of Sight) regime, the UAVs still cause transport issues as the equipment is rather heavy. The short battery capacity represents another limitation, as there may be few places to recharge without bringing a heavy power unit. Finally, operating UAVs—and their practical application—depends on good weather conditions [17].

Although not a major topic in this study, LiDAR is less affected by tree coverage and grants access to areas below the canopy. In open terrain, however, IBM may also produce high-quality 3D models, although the efficiency is not by far the same. Small lightweight UAVs (>250 g) could thus offer a simple and economic alternative with better practical applicability in remote field situations.

As to geophysical prospection, data must be collected near the surface, and access and local topography may represent severe limitations. All the sites in this study were accessible by car, which is considered a condition for most geophysical surveys. Most shielings, however, are not, and the pathways are often rough and steep. The MS2D meter is probably the only equipment that can be carried by hand along these paths. This method can access most areas because the investigations are based on point measurements. By using grids with 5 m × 5 m resolution, most sites could be covered within the half-a-day limit.

Compared to other ground-based geophysical methods, the Bartington 601 gradiometer is versatile. Nevertheless, with 0.5 m crossline spacing, this method may easily be hampered by bush vegetation and other obstacles, making data collection cumbersome. Reducing the number of sensors (from 2 to 1) increases accessibility but not necessarily efficiency. As the gradiometer is carried above ground, it is not very dependent on terrain conditions, but keeping the right pace during data collection may be challenging in rough terrain. This can be helped by putting up lines with meter markers, but this may further decrease survey efficiency.

GPR is the least effective method in these environments. Motorised GPR could only have been used on rare occasions (like at Klovset), but the carts were also hard to manage, and even low vegetation hampered the efficiency. The range is thus restricted to certain areas, typically grassy or grazed surface conditions. At Rangsetra it was not considered feasible at all.

4.2. Archaeological Results—Adequacy and Credibility

The results from the aerial surveys show how the UAV-based survey methods supplement one another well in terms of visible and less visible features. Processed DEMs and 3D models from LiDAR and IBM have proved capable of capturing and highlighting subtle features that are hardly visible to the naked eye. However, across rough terrain—such as at Rangsetra—house foundations and other features are hard to verify and distinguish from natural formations (Figure 28).

Furthermore, thermography may capture other aspects of the landscape that are not readily visible in aerial photos or DEMs, such as moisture and soil character, but the investigation clearly shows the dependency on good solar conditions and the time of day. The datasets from Klovset and Søstølen, captured after sunset, clearly exhibit the potential, but further investigations to optimise the timing are needed. Some of the same patterns may be acquired from LiDAR intensity data, but in contrast to the actual electromagnetic radiation measured by thermography, such datasets mainly exhibit the ability to absorb or reflect the same radiation (Figure 29).

Unfortunately, UAV based survey methods have not proven adequate in identifying archaeology hidden beneath the surface at the shielings, and even less visible features could easily be hampered by grass and shrub vegetation. Under favourable ground and solar conditions, such as those found at Klovset, thermal photography may highlight subsurface stone features—particularly distinct lines—but only at shallow depths or preferably with some of the stones barely visible. Despite the good conditions, neither the visual, grass-covered house foundations nor any of the deeper features, as later revealed by GPR and subsequent trenching, were detected in the thermograms.

Results from the geophysical surveys rely even more on ground conditions. Due to varying geology and vegetation, results from the magnetic susceptibility surveys appear random (cfr. Figure 4, Table 2) and with significant local deviations that are not representative of the average background measure [2] (p. 97). Generally, the best results are achieved by analysing 2SD of the mean. However, as most sites were still used in the 20th century, topsoil magnetic susceptibility mainly reflects the evidence from later centuries, as a confirmation of what we already know and what could be derived from vegetation, soil, moisture, etc. Further, it also faces challenges at abandoned sites. As clearly demonstrated at Rangsetra, the shallow penetration primarily reflects the low values in the organic deposits that covered the site after its abandonment in the Late Middle Ages, not the settlement and activity underneath. The most promising results were found at Myrset, where a belt of higher values along the edge of the site—facing the surrounding bogs—may reflect subterranean traces of iron production [53].

Whereas magnetic susceptibility measurements mainly aim at identifying and delimiting sites or areas of interest [2], gradiometry is potentially more capable of detecting archaeological features. With a 0.5 m crossline spacing, however, one should not expect features less than 1m in diameter [2,57]. However, the credibility of magnetometry in shieling landscapes dominated by metamorphic bedrocks and glacial moraine deposits is low, as the archaeological features are ephemeral and hard to separate from randomly distributed magnetic rocks [2] (pp. 102–104). At some sites, such as Klovset, the location of the magnetic deviations in relation to features identified by other means helps substantiate their archaeological significance. The same applies to Setersetran and Myrset, where magnetic contrasts could be correlated to more recent house foundations partly visible above ground. Elsewhere, the datasets were too noisy to be readily interpreted within an archaeological survey context.

The best geophysical results were obtained using GPR, and potential features were detected at all four investigated sites. Being able to study the soil in various depth slices, combined with vertical profiles or radargrams, strengthens the interpretation and its credibility [44] (p. 170–171). Even here, however, the best results were from Klovset and included a prehistoric house foundation and a stone-filled pit that were both confirmed by the following investigations.

4.3. Implementation in Developer-Led Archaeology and Cultural Heritage Management Practices

According to the Norwegian Cultural Heritage Act (NCHA) §9, archaeological surveys are conducted in order to identify sites protected by the AD 1537 boundary for automatic protection and consequently to assess if a development plan or any other measure is in violation of the law [58]. As the legal consequences for the stakeholders may be severe, developer-led surveys and assessments must provide reliable results.

Generally, good advice would be to start by employing remote sensing techniques and follow up with targeted archaeological surveys and interventions. This may, however, be a large investment without necessarily leading to any conclusive answer. This is particularly problematic in smaller survey projects at the shielings, with limited budgets, ephemeral features, and difficulties discerning ancient from recent remains. The resulting consequence may lead to this type of site being neglected due to its low visibility [11].

Less-invasive methods such as systematic soil coring, combined with small hand-dug trenches and test pits, may thus represent a better option—in terms of identifying, dating and delimiting archaeological sites. Despite technological advances, this still highlights the importance of direct physical sensing and ground truthing. Such methods could conveniently be paired with the use of aerial photographs and LiDAR data available online, as well as new images and 3D IBM based on cheap, lightweight UAVs to document the visual and experiential impression of the site at the time of the survey.

After the initial detection, however, this study has exhibited the potential for non-invasive technologies to elaborate on archaeological features and the physical content of the registered sites. In addition to mere research purposes, this could contribute to adjusting planned development and, if possible, avoid conflict with subsurface features not detected by coring. If not, non-invasive investigations may also help prepare and budget for a final rescue excavation. As the subterranean remnants at the shieling sites are somewhat ephemeral, and neither features nor archaeological deposits are covered by 30 cm of topsoil, excavations by machine have often failed to realise the potential of these early transhumant sites—throwing the baby out with the bathwater [11] (p. 55). Better knowledge in advance may thus contribute to more targeted excavations. We hope that this study may exhibit the potential and limitations of varying methods and combinations at the shieling sites, in various terrains and environments, and under various circumstances.

5. Conclusions

Our investigations have demonstrated prehistoric settlement at the shielings, partly identified by remote sensing and non-invasive methods, but also how both practical applicability and archaeological results are highly dependent on the physical nature of the sites. Transhumant shielings are not one cohesive category but cover a wide range of landscapes in terms of topography, geology, and vegetation. This requires careful consideration when assessing the potential of the different methods within cultural heritage management and developer-led archaeology.

Aerial surveys are highly effective at documenting visual landscapes and features. UAVs can travel far and are not dependent on surface conditions, but a visual line of sight is recommended. They may be hampered by logistical issues with heavy equipment in remote mountainous areas. At flat, even, and well-maintained sites, both LiDAR and IBM-based 3D models can capture subtle features that are hardly visible to the naked eye. In rough terrain, however, house foundations and other anthropogenic features are hard to distinguish from natural formations. LiDAR is the most efficient and preferred tool in areas with tree coverage, but IBM is a good option for landscape and field documentation. For features with low visibility, thermography could offer a good supplement to 3D models, aerial photos, and greyscale DEMs by emphasising electromagnetic radiation, mainly from stone features. Unfortunately, the method is highly dependent on solar conditions and vegetation cover. More systematic testing under varying conditions is needed to realise the full potential, but our study at the shielings suggests that exceptional conditions are required to detect subsurface features.

Geophysical prospection may be a better option for subterranean archaeology. However, magnetic methods are hampered by the region’s bedrock and drift geology. Hence, the results from the gradiometer surveys were noisy and contained features possibly caused by erratic magnetic rocks. Further, the topsoil magnetic susceptibility surveys suffered from low penetration, mainly reflecting patterns from the recent past or superimposed organic deposits formed after site abandonment. Although the least efficient and highly impractical, the best results were gained by GPR.

Technological advances may contribute to a better understanding of prehistoric and mediaeval shielings and pastoral mountain landscapes. However, in the context of initial archaeological surveys—particularly as part of commercial or developer-led archaeology—non-invasive methods have not yet proven capable of adequately detecting, dating, and delimiting faint archaeological sites in these remote shieling landscapes. Less invasive methods—such as systematic soil coring combined with small manual trial trenches based on direct sensing—still represent a more feasible and reliable alternative. Yet, advanced non-invasive methods may prove valuable between the initial survey and the final excavation, as they provide a better understanding of the archaeological contexts and help attain a more focused and problem-orientated investigation. However, regarding practical applicability and archaeological results, this study clearly shows the dependency on landscape affordances, soil properties, and vegetation cover in these remote landscapes. Hence, the possibilities and limitations of each method, either alone or in combination, should be considered closely, along with the type of site and environment.