Introduction

A range of different maritime archaeological sites are found in coastal areas near the shoreline. These sites are marked by strong human–environment interaction and can provide valuable insights into the history of settlement, seafaring, and maritime trade (Fediuk et al. 2018; Flemming et al. 2014; Nutley 2008; Wilken et al. 2022). However, archaeological prospection near-shore and in shallow water can be challenging due to factors like strong currents, the influence of tides, suspension of sediments and difficult access.

This paper presents new approaches to acquisition of high-resolution seismic data of a partly buried shipwreck, using a single channel sub-bottom profiler (SBP) system mounted on an Autonomous Surface Vehicle (ASV). The use of an unmanned boat in combination with an accurate Global Navigation Satellite System (GNSS) antenna and Real Time Kinematic Global Positioning System (RTK GPS) allows data to be collected in very shallow water and achieve higher density between survey lines than with traditional manual controlled vessels (Olsen et al. In review). Dense grid increases the ability to create detailed 3D models which provide a better understanding when interpreting archaeological and geological contexts. This technique is a fast, cost-effective and, most importantly, non-destructive way to investigate and monitor archaeological structures and objects buried beneath the seabed.

Ultimately, a detailed 3D model can expose essential information like horizontal and vertical distribution of the embedded remains, depth of burial, structure of the hull and construction details of the shipwreck. Advances in technology and leading-edge software for geophysical visualization are creating opportunities for further knowledge and management of cultural heritage under water.

A case study of the already known and well-documented medieval Avaldsnes ship (Fig. 1) (Alopaeus and Elvestad 2006; Wroblewski and Vivås 2018), included a combined maritime archaeological and geophysical survey in order to test these acquisition and processing methods, specially adapted to operate in arduous environments. The main objectives of this study were threefold: (1) to demonstrate the potential of an ASV equipped with acoustic sensors to detect and image the partly buried wooden hull of the Avaldsnes ship, whose position is known in shallow water, (2) to interpret the seismic records and generate a 3D model from which our knowledge of the ship can be increased, and (3) compare information gathered using conventional archaeological methods such as diving inspections, underwater test trenches and photogrammetry, with results from the geophysical investigation to highlight limitations and advantages of the different practices.

Fig. 1
figure 1

© Stavanger Maritime Museum. Used with permission

a Overview map of study area in south-western Norway with location of the Avaldsnes ship in the harbor close to the medieval church of St. Olav. Base map: Google Earth. b Sketch of the late medieval shipwreck. Illustration by H. Alopaeus. c The remains of the wreck partly buried in seabed. Photo: R. Svendsen. b and c

Full size image

The paper begins with an introduction to geophysical methods and the maritime archaeological context. Thereafter, the methods and results of the case study are focused on the main objectives, followed by a discussion on the potential and the shortcomings of the methodology.

Geophysical Methods in Maritime Archaeology

Geophysical investigations, particularly seismic imaging techniques, are effective tools for mapping and obtaining detailed descriptions of the seabed and shallow subsurface. The use of non-destructive geophysical methods has progressively been used since the early 1950s (Bjørnø 2017a, b; McCarthy et.al. 2019; Quinn 2012) and are considered well-suited for maritime archaeological surveys (Boldreel et.al 2021; Gkionis et.al. 2021; Missiaen et al. 2017; Quinn et.al 2000; Wilken et al. 2019a, b).

Underwater remote sensing systems such as side-scan sonar (SSS) and multibeam echosounder sonar (MBES) are powerful instruments for detecting shipwrecks exposed on the sea floor (Ferentinos et al. 2020; Gaida et.al 2020; Grządziel 2020, 2022). However, gathering information about shipwrecks partly buried in or covered by sediments represents a different challenge. SBP systems are widely used to identify wrecks and objects located subsurface, displaying good capabilities for mapping the geology below the seafloor and distinguish features in the shallow soils (Cvikel et. al 2017; Grøn et al. 2015; Papatheodorou et al. 2011; Plets et al. 2008b).

A number of different SBP systems are available on the market. For archaeological purposes, where capturing sufficient resolution in the upper part of the sediment layers is crucial, a chirp system which sweeps an interval of frequencies has traditionally been the preferred alternative (Grøn et al. 2007, 2015).

Modern chirp systems work well in shallow water (0.5–10 m) and may offer vertical resolution at decimetre precision in the top 20–30 m of unconsolidated sediments (Grøn and Boldreel 2014; Gutowski et al. 2002, 2015).

In recent decades, there has been a significant development in very high-resolution seismic technology as well as research into new analytical techniques. The general shift towards greater use of 3D sensors and workspaces allows for new methodology and represents a paradigm shift in the practice of maritime archaeology (McCarthy et al. 2019; Winton 2019).

Several experimental investigations with 3D SBP systems have been conducted to collect updated information on known shipwrecks (Gutowski et al. 2008; Mueller et al. 2013; Plets et al. 2009; Wilken et al. 2019a, b). These technically complex parametric systems can better explain buried archaeological structures but are highly weather-sensitive and consequently complex to use in demanding survey situations. In addition, the relatively high cost of the equipment makes it somewhat unavailable for minor project with limited funds, which is often the case for maritime environmental and archaeological research.

The possibility of generating synthetic 3D models of shipwrecks embedded in soft sediments based on 2D sub-bottom seismic sections has previously been introduced (Boldreel et al. 2017; Plets et al. 2008a), emphasizing the importance of maintaining close survey line spacing in order to achieve accurate results. The University of Stavanger has made notable progress in enhancing the ability to conduct effective surveys of dense grids. As part of their advancements, they have successfully developed an ASV equipped with a single beam chirp system (Olsen et al. In review), which is utilised and demonstrated in this study. With a state-of-the-art autonomous working platform, allowing close line-spacing, profiles recorded with off-the-shelf seismic SBP devices represent an alternative to the use of an advanced 3D system.

Case Study: The Medieval Port and Shipwreck at Avaldsnes

Avaldsnes in south-western Norway (Fig. 1a) was a centre of power from the early Bronze Age to the end of the medieval period (Skre 2018, 2020). This stable position was due to its strategic key location in the main sailing route to the north along the Norwegian coast; the ‘Northern Way’ (Old Norse Norðvegr) that has likely given the country its name. The narrow strait passing Avaldsnes was the only point along this main transportation route where people and resources could easily be controlled. South-western Norway was also central for voyages across the North Sea.

In the medieval period, Avaldsnes was the site of a fortified royal palace and a port of great importance. Indeed, it was the only medieval royal palace complex in Norway built outside a town (Hommedal 2020). Written sources and archaeological material reveal that in the thirteenth-fifteenth centuries, the port was used by merchants of the Hanseatic League (Elvestad and Opedal 2001, 2019; Ersland 2019; Fyllingsnes 2019). The port was called Nothow, probably deriving from Norw. Nautøy, ‘Cattle Island’ (Særheim 2000).

Initial mapping of the port at sea and on land started in the 1970s, with a more systematic approach around 2000 (Elvestad and Opedal 2001). Since 2019, efforts have been made to gather more knowledge about the port to answer new research questions. A maritime archaeological survey in 2020 identified the first timber remains of a medieval wharf (Reiersen and Vivås 2022). Ground penetrating radar surveys on land in 2021 revealed possible medieval structures, and a subsequent excavation confirmed the first traces of Hanseatic activity ashore (Hillesland et al. 2023).

The main source of port activity, however, is found in the cultural layers in the seabed containing a significant number of archaeological finds like ceramics, animal bones, leather and woodchips, all typical to urban contexts (Demuth 2001, 2019; Elvestad and Opedal 2019). As the extent of the cultural layers has not been established, an attempt to map these are currently ongoing, combining marine geophysical investigations with conventional core sampling (Ravnås and Fredh In prep.). The most prominent structure partly buried in the seabed is a wreck known as the ‘Avaldsnes ship’ (Fig. 1) (Alopaeus and Elvestad 2006).

Previous Knowledge of the Avaldsnes Ship

The Avaldsnes ship was a medium-sized merchant ship. The origin of its timber indicates that it was built in the region of the Vistula estuary in present-day Poland (Ditta 2021). A dendrochronological analysis shows that the ship was constructed c. AD 1395 (AD 1392–1410) and sank after a fire around 1399–1415 AD (Alopaeus and Elvestad 2006; Daly 2019). It was thus a relatively new ship when it sank. Medieval sources mention several ships sinking close to the port of Nothow (Fyllingsnes 2001, 2019).

Unlike most contemporary merchant ships, this was a fine vessel, built by highly skilled ship builders. Visual diving inspections conducted by maritime archaeologists in 2003 and 2017 revealed that the remains of the Avaldsnes ship are remarkably well-preserved. The investigations showed a clinker-built ship constructed in oak with a wide rounded bow and a narrow, straight stern. A large oak keelson incorporating the mast-step was found in the central part of the ship. Furthermore, five large riders were installed on both sides, above the framing and notching over the ceiling planks which held the keelson in place. Alopaeus and Elvestad (2006) carried out a limited excavation of the keelson and described it as a solid piece of wood with a total length of six meters, while the rectangular hole for the mast foot was nearly 0.3 m wide and 0.2 m deep.

Other details noted were animal hair and moss placed between the planks as caulking material, and wattle mats of twigs whose purpose likely has been to protect the cargo from humidity and the ship against heavy loads. Two defined heaps of stones were found within the vessel. Initially, they were considered to be ballast used to sink the ship after burning. Later observations, however, led to a new interpretation of the stone heaps as secondary installations, possibly used for mooring or as part of harbour structures (Fig. 1b, Figs. 7 and 15) (Alopaeus and Elvestad 2006; Wroblewski and Vivås 2018).

In 2004, manual drawings were made of the horizontal extent and specific construction details (Alopaeus and Elvestad 2006). The wreck was re-documented by photogrammetry in 2017 and then covered and secured with geotextile. Based on photogrammetry data, a tentative 3D-model of the preserved, submerged vessel shape was made in 2019 (Fig. 2). A reconstruction sketch of the ship was also made, drawing inspiration from contemporary iconographic and archaeological sources (Fig. 3). The figures reflect the knowledge status of the vessel shape prior to geophysical mapping, working as a reference point for interpreting the data.

Fig. 2
figure 2

The Avaldsnes ship. Photogrammetry and 3D-model of assumed shape of the preserved wreck. Only the part below the water line in Fig. 3 is preserved. Model by M. Ditta

Full size image
Fig. 3
figure 3

@ Stavanger Maitime Museum. Used with permission

The Avaldsnes ship. Drawing by H. Vatnaland, based on medieval wrecks, other reconstructions, and iconography.

Full size image

In addition to the archaeological investigations, a nearshore geophysical reconnaissance survey in the inner and outer bay at Gloppeneset (Fig. 1a) was carried out in 2002, using a small, manned vessel and a chirp sub-bottom profiler system (Grøn 2002). This survey noted some unidentified anomalies buried in the seabed but concluded that inspections by divers would be required to uncover the true nature of these possible targets. Though the shipwreck was already well-known at the time of the survey, it was not specifically mentioned.

Methods

An ASV was used for performing a high-resolution chirp SBP survey in the medieval port at Avaldsnes (Fig. 5c).

The ASV consists of a lightweight unmanned, 4.2 m long catamaran, inflatable and therefore highly portable (Minicat 2023). It protrudes only 35 cm into the water column making it ideal for navigation in shallow water. With a total carrying capacity of 450 kg, it is possible to equip the vessel with marine acoustic devices to conduct geophysical mapping in extraordinary and challenging environments.

Seismic data quality might be affected by mechanical propellors leading to bubble formation. To minimize this problem and reduce undesired noise, the vehicle is mounted with two outboard electrical engines, quiet in operation but strong enough for good manoeuvrability. Sub-bottom recording was carried out using a Meridata high-frequency single-beam chirp system (Meridata 2023), attached to the front platform by an adjustable aluminium pole (Figs. 4 and 5c). This system sweeps a frequency interval between 10 and 20 kHz, giving a vertical resolution of around 10 cm based on a quarter of a wavelength criterion with an average sound of velocity in seawater of 1500 m per second (m/s). A seismic trace was recorded every 250 milliseconds (ms) (4 pings per second).

Fig. 4
figure 4

Overview of equipment and wiring network on the ASV deck. The roof box protects sensitive electronics, the roboclaw motor controller and the on-board computer. Note the GPS antenna and chirp mounted on a pole in front of the vehicle and the two outboard electrical engines in the back. Compare with Fig. 5c (next). Design by T. M. Olsen

Full size image
Fig. 5
figure 5

a The land-based working station with computer and battery package. b Screen showing online seismic recording to the left, together with the pre-made grid for the survey to the right. c The autonomous surface vehicle used for acquisition of survey data in shallow water. d Survey sailing lines, including both lines used for mapping of sediment thickness and lines recorded with 25 cm line distance for creation of a 3D datacube of the buried Avaldsnes ship. Base map: Google Earth. Photos A–C: H. Ravnås

Full size image

In order to have a fully functioning autonomous boat and ensure accurate navigation, the vehicle is equipped with autopilot and a CPOS RTK-GPS antenna, allowing centimetre precision operated via a 4G Cloud Link. More details regarding the construction and capabilities of the ASV can be found in Olsen et al. (In review).

Mission planning and immediate data visualisation take place on  a PC at the land-based working station, where a rectangle grid was entered through a CSV file, instructing the system to where data should be gathered. Line spacing and survey speed were set to 25 cm and 1 m/s, respectively, to ensure optimal data coverage (Fig. 5).

The chirp survey was conducted on a sunny, calm summer day during ebbing tide, with the total time for acquisition of the high-resolution seismic data being 110 min (Fig. 7).

Comprehensive data sets were collected to create a 3D seismic datacube to investigate the Avaldsnes ship’s dimensions and form. In addition, some extra lines were surveyed to map the geology and sediment thickness of the area (Figs. 5d and 8).

The chirp data, recorded in an envelope format with a sampling frequency of 25 kHz, was processed using in-house scripts developed by the University of Stavanger in the Matlab and Madagascar software packages. After integrating navigational information with the time-synchronized acoustic data, standard seismic processing methods, such as tide correction, bandpass filters, and noise reduction were applied. Due to minimal wind and waves affecting the operation, processing steps such as swell and heave correction were not necessary. Although the 3D survey was designed to follow a regular set of lines, the ASV will diverge from the planned grid due to the error margin of ± 0.5 m from the autopilot GPS, resulting in erratic spaced data (Fig. 6). A gridding tool transferred the irregularly distributed traces into a rectilinear grid with a cell size of 25 cm. Integration and migration of the datasets were done using a constant sound of velocity set to 1481 m/s for seawater and an aperture equivalent to the 40 degrees beam of the high-frequency chirp system. Moreover, the data was projected to Universal Transverse Mercator (UTM) 32 N coordinates and written to the Society of Exploration Geophysists (SEG-Y) output format. The processing sequences applied are described in more detail in Olsen et al. (In review).

Fig. 6
figure 6

The navigated survey lines (yellow) illustrated together with the initially planned 25 cm grid (red). The blue line represents the boat's effective straight-line path, whereas the less successful turns are evident as wider arcs extending beyond the survey area. Figure modified from Olsen et al. (In Review) where more details regarding data collection and experiences gained during the fieldwork can be found. Base map: Google Earth (Color figure online)

Full size image

For interpretation and visualization of the SBP data, Petrel and EIVA’s Navisuite were used. To achieve a general overview of the shallow soils and geology, distinct reflectors were manually picked in the seismic profiles. The digitised horizons in the 2D sections were then exported as ASCII files and used for creating isopach maps (Fig. 8).

Furthermore, close survey lines allowed for construction of a 3D datacube with a lateral spatial resolution of 0.5 m, and a vertical resolution of 0.2 m, providing great insight into the subsurface by imaging the seismic data both vertically and horizontally in any given plane. The model was flattened on the seafloor surface and used for creating time-slices representing horizontal cuts through the seismic data volume. This approach enables the interpreter to visualize submerged objects in their original depositional context, capturing all horizons within the data cube. The technique is advantageous when mapping cultural heritage sites and allows a stratigraphic display of features with potential archaeological relevance, subsequently aiding the interpretation.

An effort was made to arrange for a full 3D reconstruction of the buried remains of the Avaldsnes ship based on the obtained acoustic data. Similar attempts have been demonstrated by Plets et al. (2008a). The embedded shipwreck is clearly recognisable as a high amplitude anomaly marked on several cross sections systematically drawn through the 3D datacube (Figs. 9 and 12).

Firstly, the seabed and the base of the high amplitude anomaly were selected on each vertical profile. Secondly, with assistance from the horizontal time-slices, the interpolated surface from these picks was used to create an isopach indicating the depth of burial followed by a 3D illustration of the submerged wooden hull and its interred details (Figs. 11 and 13). Finally, the anomaly outline interpreted from the chirp data was compared with the 3D model based on previous diving and photogrammetry documentation results (Figs. 14 and 15).

Geological and Geomorphological Setting

The study area is situated in the inner bay of Gloppeneset, where the Avaldsnes ship is found partly buried, in approximately 1 m water depth in a dynamic, intertidal environment (Fig. 1a). Today, the bay is linked to the sea via two narrow and shallow straits in the eastern and southern directions. However, the water depth in the area has changed through time and relative sea-level (RSL) curves show that the water depth was 1 to 1.6 m higher 1000 years ago (Simonsen 1978; Vasskog et al. 2019), thus more favourable sailing depths and different harbour conditions in the past. The change in water depths in the late Holocene can primarily be attributed to land uplift and gradual shallowing caused by the inflow and sedimentation of organic material.

Geologically, the area is dominated by a bedrock of quartz schist and meta sandstone (NGU 2022), which occasionally appears at the seabed and becomes exposed in the beach zone (Fig. 7). Isopach maps created of geophysical records collected in the basin, show sediments consisting of soft shallow soils with a thickness varying from 0 to 2.4 m overlaying harder deposits of coarse shell-sand. Sediment cores taken at the site confirmed the interpretation of the seismic data and identified a lithology comprising of brownish fine, sandy detritus gyttja which provides good preservation conditions for submerged cultural remains (Fig. 8) (Ravnås and Fredh In prep.). However, neither the chirp system nor the Russian corer used for sediment sampling, could penetrate the total prevalence of the hard shell-sand, and the absolute thickness of the sediments in the basin remains unknown. Further investigations are necessary to gain better control over the shell-sand’s extent, by utilizing a different SBP system, such as a boomer operating at lower frequencies, and specialised core tools designed for extremely hard sand.

Fig. 7
figure 7

Drone photography showing the surveying ASV and the Avaldsnes ship in-situ at approximately 1 m water depth. Note the piles of rocks within the partly buried ship. Outcropping bedrock of quartz schist and meta sandstone can be seen in the beach zone whilst soft, muddy sediments are clearly visible in the basin providing good preservation conditions for archaeological remains. Photo: A. Vivås

Full size image
Fig. 8
figure 8

a Isopach map showing a thickness of soft shallow soils varying from 0 to 2.4 m in the study area, with the placement of the Avaldsnes ship indicated. Diagram illustrating lithology observed in the sediment cores at test locations no.3 and no.4. b Seismic profiles used for creation of isopach map with a selected section of an example line presented in c highlighted. c 2D chirp profile acquired from NE to SW showing a strong reflector representing hard shell-sand below the soft deposits embedding the Avaldnes ship. The geophysical data corresponds well with the sediment samples. However, neither the chirp SBP system nor the sediment corer were able to penetrate the hard shell-sand’s total prevalence. Nevertheless, radiocarbon dating shows that the shell-sand is of Mesolithic/Neolithic origin. Depths in MBSL Meters below sea-level

Full size image

Nevertheless, C14 dating shows that the shell-sand is of Mesolithic/Neolithic origin, whilst macrofossils found in the softer cultural layers appears to be deposited during the fourteenth century (Ravnås and Fredh In prep.). The latter fits well with the sinking and subsequential burial of the medieval Avaldsnes ship.

Moreover, underwater trenches dug by maritime archaeologists, show that the embedded Avaldsnes ship probably has acted as a local sediment trap allowing the wreck to be filled with fine-grained clay deposits, lower in organic content than the surrounding sediments (Wroblewski and Vivås 2018).

Observations and Results

Following the main objectives, the focus was on detecting the Avaldsnes ship, producing a 3D reconstruction, and compare the newly obtained data with existing documentation.

To secure the site and avoid extensive erosion and rapid degrading of the Avaldsnes ship, after the initial archaeological investigations, the exposed frame of the ship was covered with geotextile, held down by several sandbags (Fig. 15b). In-situ preservation is vital, either as a permanent measure, or until a potential excavation or raise of the wreck may take place (Bjørdal et al. 2012; Maarleveld et. al. 2013). However, these protective artefacts represent a source of undesired noise while acquiring high-resolution geophysical data. The sandbags attenuate energy and prevent chirp signals from penetrating below the structures, leading to an acoustic blanking zone in the seismic records. The stone piles within the wreck amplify this distraction (Fig. 9).

Fig. 9
figure 9

Depth slice revealing the wreck outline at − 1.02 m with example profiles of composite lines a to d highlighted. Note the piles of rocks within the ship. Each profile a–d is presented with and without interpretation. The 2D sections show the stone heaps and sandbags installed at seabed leading to an acoustic blanking zone in the seismic records. A high amplitude anomaly is observed with varying continuity representing the buried hull. All depths in MBSL Meters below sea-level. See Fig. 12 for all composite lines (Color figure online)

Full size image

Despite these issues, the wooden hull of the shipwreck was easily distinguished in the seismic profiles as a synclinal-shaped high amplitude anomaly, which truncates sub-horizontal reflectors representing the local stratigraphy in the 2D sections (Fig. 9). The high amplitude anomaly was observed with varying continuity in a north-westerly direction evident from the seafloor, extending to a maximum depth of 0.8 m, attributed to be a part of the keel.

Moreover, further analyses of the vertical profiles revealed an, until now, undiscovered feature. In close proximity to the stern, protruding towards the southeast a prominent elongated target with dimensions 1.2 m × 2.0 m × 0.3 m is discernible in the sediments at 0. 8 m depth. Its location, in relation to the wreck site, makes it reasonable to suppose that this might represent parts of the rudder or other wreck-related structural elements of archaeological importance (Fig. 13).

In addition to the digitalisation of distinguished reflectors in the profiles, horizontal sections, or time-slices originating from the 3D cube, proved to be extremely useful for interpretating the high amplitude anomaly appearing as a characteristic ovate planform pattern coinciding with the ship location (Figs. 10 and 15c). These reflection events concentrate at the first meter below the seafloor, showing the installed stone heaps and sandbags on top of the structure followed by the distinct wreck outline observed downwards as it narrows in and disappear completely at a depth of 0.8 m.

Fig. 10
figure 10

a Uninterpreted time slices of the 3D datacube every 12 cm downwards showing the major reflection features of the shipwreck. Note the rocks and the sandbags starting to be prominent already at a depth of − 0.68 m. b Interpretation added to the time slices. c The 3D cube has been flattened on the seafloor surface. All depths in MBSL Meters below sea-level

Full size image

The time-slices provided clear images of the subsurface contributing to enhanced extraction of accurate dimensions, shape, and location of the buried shipwreck remnants. Figure 10 shows a defined and pronounced longitudinal and axial symmetry of the Avaldsnes ship with a length of 18 m, a width of 5.6 m and a maximum burial depth of 0.8 m. Dimensions derived from the 3D cube and seismic profiles correspond well with the measurements reported during diving operations, although the maximum depth of burial demonstrates some deviation from previous investigations (Table 1).

Table 1 Comparison of the Avaldsnes ship’s dimensions from diving and geophysical studies
Full size table

The noticeable deviation observed in depth may be related to the constant sound of velocity set to 1481 m/s during data processing and conversion from seismic two-way travel time. This velocity appears to be too low for the near-surface sediments. Using a higher velocity for the seismic events below the seabed would lead to an increase in depth.

Another explanation may be that the high amplitude anomaly interpreted in the 2D sections, represents the upper part of the ship remains and it must be considered that the hull has several layers and components, some of which have an unknown extent (hay, mats of wattled twigs, ceiling planking, keel, frames, and outer planking). The most regular and structurally solid feature is the oak keel, which archaeologists estimated to be up to 0.5 m thick, 0.20 m wide and measured to be 17.6 m long.

Furthermore, the newly obtained seismic data not only enhanced our understanding of the Avaldsnes ship and possible wreckage, but also unveiled additional, previously unknown, anomalies in the bay’s eastern section. Several relatively large-scale objects with high acoustic reflectivity were imaged on the amplitude maps. These could potentially represent parts of the foundations of a ship dock from the same historical period. However, these anomalies are yet to be verified by traditional archaeological methods.

The interpretation of the shipwreck remains, from all acoustic profiles (Figs. 11 and 12), were arranged to create a georeferenced 3D reconstruction of the ship and the potential wreck site extent (Fig. 13). Isolated small parts and minor construction elements such as the frames of the ship could not be correlated with seismic reflections or diffractions. This is possibly due to missing cross lines covering the survey area, or that the interpolation method used, has smoothed out certain details important for recognizing the frames, or inner planking. Nevertheless, the 3D reconstruction reveals precisely positioned information about an approximately 6 m long feature in the middle of the structure, believed to represent the keelson and raiders. Assuming the high amplitude anomaly observed in this section is the upper part of the keel, the position of the potential keelson within the shipwreck corresponds to previous archaeological documentation (Figs. 14 and 15a). Further, the shape and location of the detectable asset match the keelson/mast-step and its dimensions are close to what previous investigations reported (Fig. 2).

Fig. 11
figure 11

Isopach map showing the thickness of the buried Avaldsnes ship remains (in meters below seabed). Illustration based on interpretation of the high amplitude anomaly observed in the SBP data. Data indicates that the remaining hull and keel is tilting slightly towards southwest. Coordinates refer to WGS84 UTM32N

Full size image
Fig. 12
figure 12

Amplitude map at − 1.04 m depth showing the contour of the Avaldsnes ship. The coloured lines on top represent the vertical cross sections, systematically drawn trough the 3D datacube in various angles

Full size image
Fig. 13
figure 13

Interpretation from the acoustic 2D profiles arranged to visualise a georeferenced 3D reconstruction of the Avaldsnes ship and the potential wreck site extent with a vertical exaggeration 1.5x. White lines = wooden hull, red lines = keelson + raiders, green dots = object (1.2 m × 2.0 m × 0.3 m). Object possible represents parts of the rudder or other wreck-related structural elements of archaeological importance a 3D model seen from south-north. b 3D model seen from north–south

Full size image
Fig. 14
figure 14

a True to scale cross section of the Avaldsnes ship visualised on a 2D seismic profile cutting through the middle of the ship (composite line 98—position seen in b). Internal details like the keelson and mast step correspond well with the high amplitude reflectors, supporting the observation of a slight tilt towards southwest b Composite line 98 cutting through the Avaldsnes ship and keelson c Photogrammetry detail of keelson by M. Ditta

Full size image
Fig. 15
figure 15

Results from the latest diving campaign and cross section trench compared with the geophysical survey and recent drone photography. a Site plan and documentation of observations from diving operations. Scale 1:50. Drawing: M. Ditta. b Drone photo of the buried Avaldsnes ship showing details like stone heaps and installed sandbags. Photo: A.Vivås c Drawing of the Avaldsnes ship laid on top of a georeferenced seismic timeslice showing good correspondence in the ship’s dimensions and shape

Full size image

Although individual details of the oak framework are difficult to discern, the shipwreck representation provides a well-defined model in relation to other artefacts and the surrounding environment, revealing a slight tilt of the wooden hull towards the southwest (Figs. 11 and. 14a). The model answers basic archaeological questions such as spatial delimitation of the wreck structure and the wreck site, important knowledge for the management of the heritage site and potential future excavations.

Discussion on Methodology

Let us then turn to the potential and the shortcomings of the presented methodology.

Ongoing research and developments in the field of robotics represent new methodological opportunities for maritime archaeological investigations, especially in challenging shallow water environments (Gasperini et al. 2014; Olsen et al. In review; Stanghellini et al. 2020). In this case study, employing a customized ASV, it was possible to acquire close-spaced high-quality 2D sub-bottom chirp data in the medieval port of Avaldsnes. The geophysical reconnaissance survey, carried out by Grøn (2002), utilized a small, manned boat, producing data of limited quality where the ship was not identified. Using a manually driven vessel in intertidal environments gives restricted steering capabilities and manoeuvrability, which can lead to inaccurate positioning, and the noise from the mechanical propeller may affect the acoustic data quality (Brennen 2005). Furthermore, in contrast to the data that forms the basis for this paper, the mapping in 2002 operated with a line distance of 20 m, which means significantly less accuracy and a greater risk of overlooking anomalies which could represent features of archaeological importance. Moreover, a close-spaced grid of high-resolution seismic reflection lines is essential to create informative 3D models to better understand archaeological and geological contexts.

Due to the very shallow water, the time allocated for data acquisition was limited by the tide window. The choice of a 25 cm sampling interval, combined with an ASV cruise speed of 1 m/s, was optimal as it allowed for the most consistent sampling while adhering to the fieldwork constraints. Achieving a higher sampling resolution could be feasible by elevating the ping rate and increasing the density of transects. However, higher ping rate may introduce more noise in the water column, stemming from reverberations between the shallow seabed and the sea surface. Furthermore, sailing additional transects would extend the survey duration and necessitate the replacement of extra boat batteries.

Conducting non-destructive mapping of the Avaldsnes ship which is found at a water depth less than one meter, made it impossible to lower the chirp to reduce the distance between the sensor and seabed. Nevertheless, given the superficial nature of the target, making such an adjustment would probably result in minimal impact of the data resolution.

Olsen et.al (In review) performed theoretical and numerical analysis of the resolution of the chirp system used. Based on the observations in their results, objects around 50 cm in size are discernible at a depth of 1.5 m.

An alternative approach and improved strategy to achieve better resolution would involve utilizing a 3D SBP profiler, capable of enhancing sampling density without extending the survey duration. Conversely, as previously noted in the introduction, it poses a challenge that this type of equipment is often relatively expensive and, in many cases, surpasses the financial capacity of a project with restricted funding.

Commercial high-frequency instruments designed for near-surface applications are typically equipped with their own established processing and imaging software that rely on fundamental operations. However, advanced refinement steps like deconvolution and migration are frequently not applied in shallow soil surveys (Plets et al. 2013). This study used 3D interpolation and migration techniques to appropriately reposition dipping reflectors, concentrate scattered energy from specific targets, and ensure an evenly sampled data volume in all three dimensions. To produce such 3D representations requires some degree of sophistication in computing power and often also specialised knowledge of specific computer programs. For small-scale survey projects this is not desirable. The aim should thus be a transition to inexpensive open-source software packages to ensure that the workflow is accessible for projects with limited budgets and resources.

Another issue to be considered for future surveys is parameters and settings used during data acquisition. The high-resolution chirp data was recorded in the envelope of the signal. A fullwave format would possibly allow for better resolution of the stratigraphy and enable the polarity of the signal to be used to perform acoustic impedance inversion procedures. The acoustic impedance could be useful to distinguish between different material and lithologies of the survey area. Shortcomings and suggestions for future improvements of the methodology in both data acquisition and processing is explored in greater detail in Olsen et al. (In review).

Regarding detailed mapping of underwater archaeological sites, two main methods are often employed: visual inspection performed by divers and acoustic surveys. While both methods aim to uncover information, they differ in their approach and the documentation they can provide. Visual inspection relies on the ability of divers to physically explore the site and make observations based on what they see, while acoustic surveys use remote sensing technology to gather data about the site's topography and possible buried features. This work delved into the main differences between these two methods while examining their respective strengths and limitations in mapping the medieval Avaldsnes ship whose location is known.

In this case study we were in a unique position considering the access to extensive information from previous maritime archaeological investigations. Divers have visually inspected the Avaldsnes ship and wreck site through several field seasons. As a result of physical measurements, comprehensive logging, and documentation from underwater test trenches together with photogrammetry, a significant base of knowledge has been built up.

The technique of photogrammetry is often used in marine archaeology to create 3D models of underwater features and artefacts, including shipwrecks (Aragón et al. 2018; Balletti et al. 2015). However, this approach is dependent on good visibility and access to achieve optimal angles of the object of interest. Even though the process of photogrammetry is a non-invasive method, digging of trenches to prepare for inspection in this case was a necessary intervention to expose the partly buried ship remains for documentation.

Photogrammetry combined with professional knowledge of ship building traditions, documentation and experience from other comparable medieval ships made it possible for archaeologists to create a tentative 3D model of the preserved, submerged vessel shape and construction details.

On the contrary, high-resolution chirp data has the advantage of being able to penetrate through sediment and other obstructions, providing a view of the buried artifact or structure that might otherwise be impossible to obtain without excavations. However, all geophysical methods are inherently uncertain and provide only indirect information, and accuracy of interpretation and the consecutive model depends highly on the quality of the data. Non-destructive geophysical methods in general can identify undiscovered wrecks and gain new knowledge about known wrecks prior to archaeological excavation.

Although previous studies provided significant information about the wreck, there were still unanswered questions. In particular, the dimensions and shape of the deeper buried parts of the Avaldsnes ship and the design of the keel were poorly understood. The presence of possible wreck-related objects buried in the sediments close to the ship were also unknown prior to the present study.

While the visual inspections performed by divers revealed small details like nails used in the ship construction and minor items such as ceramic shards, ropeworks and leather fragments, these were impossible to discern in the acoustic data. Nor could the presence of charred ship parts, from when the ship burnt before sinking, be observed in the dataset. On the other hand, the seismic records contributed with new and complementary knowledge about the Avaldsnes ship by implying a slight tilt of the hull towards the port side and a so far unreported embedded object in the vicinity of the ship’s stern.

The two different survey methods allowed for creation of 3D models from their respective data sets. Comparing the two 3D models shows a good correspondence of the ship dimensions, even though a divergence in maximum depth of burial is debatable. Nevertheless, larger structural features like the keelson are recognizable in both models and supports that previous observations can be decipherable in high-resolution chirp 3D datasets.

Through data integration and cross-examination, combined with ground-truthing outcomes, the datasets act as interpretive and complimentary to each other, and their synthesis provides a comprehensive digital picture of the shipwreck.

Conclusions

Acquisition of geophysical data in shallow water is challenging, and the development of expertise and experience in this discipline will fill a professional and commercial void. This paper has successfully demonstrated the use of a single-channel chirp system mounted on an ASV for detecting and mapping buried archaeological objects in a dynamic, intertidal environment. The ASV provides fast mobilization and easy access to areas where it usually would be difficult to steer traditional boats. Operating an unmanned working platform allows for a very time-efficient, non-invasively study of the seafloor and the subsurface features by collecting high-resolution seismic data in a dense grid.

All three objectives of the case study were successfully accomplished. (1) The embedded Avaldsnes ship was detected and imaged using an ASV equipped with SBP and reliable positioning sensors. (2) By utilizing the combination of acoustic vertical sections and horizontal amplitude maps, the outline of the wooden hull could be accurately identified both in plan-view and in 2D seismic sections. These findings ultimately led to a useful 3D reconstruction, which provided valuable insights into the vessel’s dimensions and maximum depth of burial. The results of the reconstruction supported the previous diving investigations and further contributed to our understanding of the ship's buried structures, disclosing a slight tilt of the keel and a large object buried in the sediments, that is probably part of the rudder or other wreck-related remains of archaeological importance. (3) Finally, the study compared the information gathered using conventional archaeological methods such as diving inspections, documentation from underwater trenches and photogrammetry with the results from the geophysical investigation. By discussing and highlighting the limitations and advantages of these different practices, the study provided valuable insights into the strengths and weaknesses of each method, paving the way for future advancements in underwater archaeological research. While the case study indicates that a multi-methods approach is favourable, it also suggests that identification of unknown submerged wrecks in the future might be possible using solely geophysical methods.