Abstract
Two unique Pre-Viking Age ship burials were found from Salme village, Saaremaa Island, eastern Baltic Sea, containing remains of seven men in the smaller and 34 men in the larger ship. According to the archaeological interpretations, these ships belonged to a viking crew possibly from the Stockholm-Mälaren region, eastern Sweden. Geoarchaeological research was conducted in the area to reconstruct Late-Holocene relative sea-level (RSL) changes and shoreline displacement to provide environmental context to these burials. In this paper we present a Late-Holocene shore displacement curve for the Saaremaa Island and GIS-based palaeogeographic reconstructions for the Salme area. The curve shows an almost linear RSL fall from 5.5 to 0.8 m a.s.l. between 1000 BC and 1300 AD with an average rate of 2 mm/year. A slowdown in regression may be attributed to accelerated sea-level rise after the Little Ice Age and during the industrial period, being consistent with the tide-gauge measurements from the 20th century. Palaeogeographic reconstructions indicate the existence of a strait in the Salme area during the burial of the ships. The eastern part of the strait with water depth up to 2.8 m was about 80–100 m wide. The relatively steep and wind-protected shores in that part of the strait were probably the best places in the area for landing the viking ships. According to sedimentological evidence and diatom data, the narrowing of Salme palaeostrait occurred between 1270 and 1300 AD. Salme I and II ships were buried at 650–770 AD into the sandy-gravelly coastal deposits which had accumulated there in the open coastal zone about 710–450 years earlier. Reconstructions show that the ships were located about 2–2.5 m above coeval sea level and more than 100 m from the coastline. Thus, both ships were probably moved from the shore to the higher ground for burial.
Introduction
The Salme ship burials form a unique archaeological find complex, one which has provided a great deal of substantial data about the Scandinavian Vikings and their expeditions (Margaryan et al., 2020; Price, 2020; Price et al., 2020). There are very few burials of Viking war bands that are known in the world. Salme is the oldest one of them, and the only one where fallen warriors have been placed together in their ships.
The site which consisted of two vessels is located on the Estonian island of Saaremaa, in the village of Salme, where it was uncovered in 2008. Seven men were buried in the smaller vessel (Salme I) and 34 in the larger one (Salme II). In addition, weapons, gaming pieces, combs and other equipment along with the provisions which would be needed for the voyage, and also some sacrificed dogs and falcons were deposited as grave goods inside the ships (Konsa et al., 2009; Peets et al., 2011, 2013). Radiocarbon dates from the human bones and typo-chronology for artefacts tend to place the Salme ship burials between 650 and 770 AD, to the end of what is known in Estonia as the Pre-Viking Age, and in Sweden as the Vendel Period (Konsa et al., 2009; Luik et al., 2020; Margaryan et al., 2020; Peets et al., 2011).
According to the isotopic analysis, the individuals who were buried with the ships at Salme were of a non-local provenance. Five of them probably spent their early life in Gotland, but the majority came from eastern-central Sweden (Price et al., 2020). Comparative analysis of artefacts and burial rituals also suggests that these men had travelled to Saaremaa from the Mälaren region in Sweden (Konsa et al., 2009; Luik et al., 2020; Peets and Maldre, 2010; Peets et al., 2011, 2013; Price et al., 2016). Numerous ante-mortem injuries from sharp-edged weapons on the men’s bones indicate that they died violently in armed conflict (Peets et al., 2013; Price et al., 2020).
The crews were probably attacked on their way along one of the main trading routes across the Baltic Sea, one which ran across the archipelago of the Åland islands towards the amber-rich Courland region of Latvia (Price et al., 2016). The shortest route from Courland to the Mälaren district in Sweden possibly ran through the Salme Strait, which existed in place of the present-day shallow, narrow River Salme (Figure 1). Due to the postglacial apparent land uplift, which around this area occurs at a rate of about 2 mm/year, the strait became narrower and turned into a river. However, the configuration and depth of the strait, the relative sea level (RSL) during the Vendel Period, and main archaeological uncertainty about the location of the ship burials in relation to the contemporary coastline – whether it was submerged or above the waterline at the time – still remains unclear.
(a) An overview map of the Baltic Sea basin area with present-day absolute land uplift isobases (mm/year, after Vestøl et al., 2019) and Viking Age transportation network (dashed line after Mägi, 2015), (b) an overview map of Saaremaa Island with locations of the sites mentioned in the paper and (c) the study area shown on the Estonian Relief Map (Estonian Land Board, 2020), with locations of new coring sites, geological and GPR profiles and Salme ship burials (ships not in scale).
The existence of the Salme Strait can also be spotted on some of the oldest maps of the area which date to the beginning of the 18th century. Unfortunately, the maps of that time are geometrically unreliable, unequally distorting various elements in the landscape. To be able to clarify the palaeogeography of the Salme Strait and the RSL changes in the area, geoarchaeological fieldworks were carried out in the Salme area between the years 2015 and 2020.
The aim of this study was to reconstruct Late-Holocene RSL changes and palaeogeography of the Saaremaa and Salme area in order to test the hypothesis regarding the possible existence of the Salme palaeostrait and to reconstruct its depth and configuration during the Vendel Period. Our study contributes to earlier knowledge by providing new AMS-14C and OSL dates, a new diatom record, scanning electron microscope (SEM) analysis of mineral particles, GIS-based palaeogeographic reconstructions and a Late-Holocene sea-level curve for this region. The study provides new RSL indicators and sheds a new light on Late-Holocene RSL variability which has not previously been studied on Saaremaa Island.
Geological setting and archaeological background
Saaremaa Island, which forms the main barrier between the Gulf of Riga and the Baltic Sea, is the largest island in Estonia and the fourth largest island in the Baltic Sea (Figure 1a). The island’s main landforms are inherited from the last glaciation and have been later reworked by the Baltic Sea, which is almost non-tidal (Leppäranta and Myrberg, 2009). The area is characterised by a slow postglacial isostatic rebound with absolute uplift relative to the geoid of about 2.3 mm/year (Suursaar et al., 2019; Vestøl et al., 2019).
Our study area is located on Saaremaa Island, on both sides of the River Salme, which separates the Sõrve Peninsula from the Saaremaa mainland (Figure 1). The river is a remnant of the former Salme Strait which itself was part of a trade route and a passage to settlements on both sides of Saaremaa Island. The strait was actively used before and during the Viking Age. Nowadays only a small river remains of the strait, having been straightened and deepened in most parts.
The easternmost part of the River Salme flows through an old sandy-gravelly coastal spit which stands out in the relief, being higher than the surroundings (Figure 1c). The sandy-gravelly sediments lie on bluish-grey layered silty clay (varved clay) which in turn lays on glacial till with an upper surface ca. 13–16 m below the ground surface (Metsküla et al., 1984). On the eastern side, the spit is abruptly bordered by the sea, whilst to the west, it slowly slopes towards the River Salme. The river channel has been cut into the clay (Metsküla et al., 1984). Most of the Salme palaeostrait area can now be described as a coastal plain with altitudes of between 0.5 and 2.0 m a.s.l. The palaeostrait widens westwards to more than 2 km and the former coastline can be followed in the relief at an altitude of ca. 3.5 m (Nirgi et al., 2017). Our study area covers the narrowest (ca. 60–240 m in width) part of the palaeostrait (Figure 1c).
Salme ship burials were located north of River Salme and 230–270 m west of the present coastline (Figure 1c). The bottom of the Ship I was 4.1 m above the present-day sea level, and the bottom of Ship II was 3.7 m above. The ships were 40 m apart, and both were similarly orientated in the north-eastern-to-south-western direction. The upper areas of the clinker-built ships had not been preserved and the grave mounds were absent, which is otherwise rather common with this form of burial sites. Based on preliminary data, researchers have previously suggested that the ships had been dragged about 100 m inland and had partially been covered with stones and soil (Price et al., 2016). Then they were completely buried quite quickly by marine sediments which were washed ashore by storms.
The Salme burial complex is unique across the whole of Europe, being distinguished from other ship burials of a similar date and with analogous find material by a large number of the buried and the singleness of the burial practices and the rituals this tends to reflect (Peets et al., 2013).
Material and methods
Sediment sampling and chronology
Nine trenches were dug on a profile which was perpendicular to the coastline in order to investigate the sedimentary structures of the coastal landform system, and to get samples for luminescence dating (Figure 1c). The OSL samples were collected from undisturbed sandy-gravelly sediments using opaque plastic tubes. Three samples from three trial pits were analysed at the Lund Luminescence Laboratory, Sweden. After wet-sieving to extract the selected grain-size fractions, preparation included treatment with 10% HCl to remove carbonates, 10% H2O2 to remove organic material, and density separation at 2.62 g/cm3 (LST Fastfloat). The quartz separate was then treated with 38% HF to remove any remaining impurities and to etch the outer surface of the grains, and finally with 10% HCl to clean out fluorides. Water content was considered individually for each sample (Table 1). The sediment dose rate was determined by high-resolution gamma spectrometry at the Nordic Laboratory for Luminescence Dating, Aarhus University, Denmark, and the total environmental dose rate was calculated in the DRAC online calculator (Durcan et al., 2015). Large (8-mm) single aliquots of 180–250 μm quartz were analysed in a Risø TL/OSL reader, model DA-20 (Bøtter-Jensen et al., 2010). Pre-heat plateau and dose recovery tests were carried out in order to adapt the analytical protocol to suit the samples. A ‘Single Aliquot Regeneration’ (SAR) protocol with post-IR blue stimulation was used (Banerjee et al., 2001; Murray and Wintle, 2000, 2003), with preheat and cutheat temperatures of 240°C and 200°C. OSL stimulation was by blue light sources (470 ± 30 nm; ~50 mW/cm2), and detection was through 7 mm of U340 glass filter. Doses for age calculation were calculated using the arithmetic mean (Guérin et al., 2017) and the ‘Central Age Model’ (CAM; Galbraith et al., 1999). The OSL ages were finally recalculated into BC/AD dates by subtracting the OSL age from the year of sampling (2015); ages and errors were rounded up or down to the nearest 10 years.
Summary of results of the quartz OSL ages.
Preferred ages are marked in bold.
CAM: central age model; n: number of aliquots; w.c.: water content.
Age BC.
Sediments of the Salme palaeostrait were described in 32 cores along three profiles (Figures 1c and 2a). Coring was carried out using a 1-m-long Russian-type peat-corer (inner ø 50 mm). Altitudes for the trial pits and coring sites were determined using the LiDAR digital elevation model (Estonian Land Board, 2020). All the altitudes are given in the European height system (EVRS, Amsterdam zero). A 2.5-m-long master core was obtained from the bank of the River Salme at coring site SV02 (Figures 1c and 2b) so that samples could be taken for sedimentological and biostratigraphical analyses, and for age determination. Additional data on the spatial distribution of the sediments in the Salme area were incorporated from geotechnical studies by Metsküla et al. (1984) and Eller and Sedman (1985).
(a)-(c) GPR sections allow distinguishing two sandy-gravelly beach sediment units (separated by a green dotted line) that lay on top of varved clay (black dotted line). The ships are buried in the upper unit. The groundwater level is marked with a blue line and approximate positions of the dated samples are marked with a red dot. Note the difference in horizontal scales. The locations of the profiles are shown on Figure 1c.
The seeds of terrestrial plants and some pieces of charcoal were collected from the Salme master core for AMS (Accelerator Mass Spectrometry) radiocarbon dating. Eight samples from the master core and one additional sample of a shell fragment from the coastal deposit (same location as OSL-sample no. 18003) were analysed at the Tandem Laboratory in Sweden. The radiocarbon ages, including earlier datings from archaeological sites, were converted to calibrated ages (BC/AD, within a 2 sigma deviation – probability 95.4%) using the IntCal 13 calibration curve (Reimer et al., 2013) in the OxCal v4.2.3 software (Bronk Ramsey, 2009). The same program (with P_sequence option) was used to create an age-depth model to assess the sediment accumulation rate for the sediment sequence in the palaeostrait. RSL data for the area were collected into a unified database based on the HOLSEA format (Hijma et al., 2015; Supplemental Material, available online; Table 3) and were used to reconstruct the RSL curve for the area (Figure 8).
Summary of the sea-level data presented in this paper. Full data set is available online, in the Supplemental Tables.
SLIP: sea level index point; ML: marine limiting point; TL: terrestrial limiting point; NS: not suitable as a limiting point.
Age BC.
Out of the time-range of the RSL curve in this study.
The content of organic matter and carbonates in the sediment samples was determined by loss on ignition (LOI), following the methodology described by Sutherland (1998) and Heiri et al. (2001). Grain size distribution in the mineral components of the dated palaeostrait sediments was analysed using the Mastersizer 3000 laser diffraction particle size analyser (Malvern Instruments Ltd, 2013). Prior to measurement, the samples were dispersed using ultrasonic action. The degree of sorting and grain sizes for each sample were classified following the methodology provided by Folk and Ward (1957). Grain characteristics such as mean grain size and sorting were analysed using the computer program, GRADISTAT (Blott and Pye, 2001).
Diatom analysis
Diatom samples were prepared following standard techniques (Battarbee et al., 2001). A drop of the cleaned residue was spread over a coverslip, dried overnight at room temperature, and fixed in Naphrax™ (refractive index – 1.73) onto a microscope slide. The diatom identification was carried out under oil immersion differential interface contrast 1000× magnification, using a Zeiss Axiophot light microscope. At least 400 diatom valves were counted and identified to species level for each sample. The diatom taxonomy and information about their environmental preferences and habitat were obtained from the Baltic Sea inter-calibration guides (Snoeijs, 1993; Snoeijs and Balashova, 1998; Snoeijs and Kasperovičienė, 1996; Snoeijs and Potapova, 1995; Snoeijs and Vilbaste, 1994), as well as other well-established diatom floras (Krammer and Lange-Bertalot, 1986, 1988, 1991a, 1991b; Lange-Bertalot et al., 2017; Witkowski, 1994; Witkowski et al., 2000). Diatoms were divided into groups according to their salinity tolerance: marine/brackish/halophilous, indifferent, small fragilarioid taxa with brackish-water affinity, small fragilarioid taxa preferring freshwater, and freshwater taxa. Habitat classification included planktonic, small fragilarioid, and periphytic taxa. A diatom diagram was drawn by using the computer program, Tilia v.1.7.16 (Grimm, 2011), along with CorelDRAW X6.
Scanning electron microscope (SEM)
SEM analysis of mineral particles (quartz grains) makes it possible to detect the type of sedimentary environment in which the grain has stayed, along with its post-sedimentary transformation (Costa et al., 2013; Kasper-Zubillaga, 2009; Krinsley, 1980; Krinsley and Trusty, 1985; Mahaney, 2002; Mahaney et al., 2010; Smith et al., 2018). Eleven sediment samples were subjected to the analysis from the master profile which corresponded either to the thin alternating silt/sand and gyttja layers (seven samples from 30 to 150 cm), to the uppermost layer of organic rich silt and soil (two samples at 10 and 30 cm) and to the lowermost sand and organic-rich silt (two samples at 190 and 200 cm). Prior to analysis, the sand sediment fraction was retained from organic matter through wet sieving with the use of a 63 µm sieve. About 10% HCl was used to remove any carbonates from the samples, followed by a few cycles of rinsing with distilled water. Samples were oven-dried and grains were made ready to place onto double-sided tape on an SEM stub. Because it was mainly a fine-grained sand fraction which was now prevalent, a tiny amount of sieved sediment was placed onto tape with the use of a spatula. Grains were randomly chosen and 20 grains per sample were processed following the recommendation by Costa et al. (2012) and Vos et al. (2014), resulting in 220 analysed grains (11 samples in profile). A Zeiss EVO MA 15 at the Department of Geology, University of Tartu, Estonia, was used at a magnification between ca. 500-times and 4000-times in order to determine a general grain outline and its surface details, namely microtextures, following the atlas by Mahaney (2002) and the review by Vos et al. (2014). A total of 36 microtextures of the mechanical, chemical, and combined type were checked out on every single analysed grain. Semi-quantitive environmental-wise discrimination was used, as offered by Vos et al. (2014), which considers an occurrence of microtextures and with some modification. Microtexture occurrence was grouped as ‘abundant’ (while occurring on ⩾75% grains), ‘common’ (50–74%), ‘moderate’ (21–49%), ‘sparse’ (6–20%), ‘rare’ (<5%) and ‘not observed’ (0%). Here, one more occurrence group (‘moderate’) following Kalińska-Nartiša et al. (2018b) was introduced comparing with an original work by Vos et al. (2014).
One reference grain sample from the coastal beach-ridge which corresponds to the 18001 OSL-dated sample (Table 1) was also investigated under SEM. However, a coarser and dominant quartz fraction of 0.5–1.0 mm was considered, and magnification between 100- and 2000-times was used.
Ground-penetrating radar analysis
Ground-penetrating radar (GPR) was used to gain more information about the bedding and distribution of sediments in the study area. A Zond-12e radar device (Radar Systems Inc., Latvia) with a 300 MHz antenna was used. The GPR antenna was pulled along at walking speed and measurements were triggered at a spacing of 5 cm by odometer wheel. Measurements were coordinated with the hand-held GPS device (with a positional accuracy of 5–7 m based on a comparison with the orthophoto), which was connected to the GPR device but, in addition to the drilling sites, ship locations and other details were marked out on profiles.
GPR data were processed using Prism2 software. Processing included minor frequency filtering, gain adjustments, time-to-depth conversion, and the application of topographic correction (based on LiDAR-DTM by Estonian Land Board). Groundwater level was usually well visible in sandy-gravelly sediments. Since water saturation tends to significantly change electromagnetic wave velocity, the time-to-depth conversion was performed using relative permittivity values of 8 and 16 respectively for sediments above and below the water table. The thickness of sandy sediments – that is, the depth to varved clay – was verified against geotechnical drilling data (Eller and Sedman, 1985; Metsküla et al., 1984).
RSL reconstruction
The data were divided into three groups according to their indicative meaning: (1) sea-level index points; (2) marine limiting data; and (3) terrestrial limiting data. Indicative meaning reflects the position of the RSL in relation to the data point at its time of deposition (Hijma et al., 2015; Khan et al., 2019). Therefore, for example, the dates from archaeological settlements were used as terrestrial limiting data, but data points from marine sediments were used as lower limiters. The curve was roughly drawn according to the RSL data, whereas the vertical values were calculated by a formula in the HOLSEA database, which considered the altitudes of the samples together with given errors.
Desktop mapping for the highest shorelines, and palaeogeographic methods
The mean elevations of the two highest shorelines which are traceable around the Salme palaeostrait system were calculated based on the methodology which was described by Rosentau et al. (2020) and the LiDAR-DTM (Estonian Land Board, 2020). The elevation was measured at every 40–60 m along the shoreline in the respective DTM. The elevations were sampled only in locations in which the foot of the ridge or escarpment which indicated the position of the palaeoshoreline was visible in the current landscape. The measured elevations, for example, the mean elevations are given in Table 2.
The mean elevations of the highest shorelines around the Salme palaeostrait, calculated based on the terrain model.
The palaeogeographic reconstructions for Saaremaa Island and the Salme Strait are based on the GIS approach (Rosentau et al., 2011), whereby the palaeo-water-level surfaces were subtracted from the 1 × 1 m resolution LiDAR-DTM (Estonian Land Board, 2020). In order to create accurate palaeoreconstructions, the younger sediments (with max thickness of 2.1 m) in the narrowest part of the Salme Strait area were subtracted from the LiDAR-derived DEM according to data by Eller and Sedman (1985), Metsküla et al. (1984) and AMS radiocarbon dating and sediment thickness data from the present study. The palaeo-water-level surfaces were interpolated using a point-kriging approach between the Litorina Sea and modern Baltic Sea water-level surfaces. The Litorina Sea water-level surface has been compiled according to the databases which cover Litorina Sea coastal formations (Saarse et al., 2003), and there is a vertical error of ±1 m. The interpolated water-level surface for the modern Baltic Sea is based on sea-level measurements (1892–1991 AD), complemented by geodetic data (Ekman, 1996). The Bronze Age and Viking Age time slices which are presented in this paper were interpolated between these reference surfaces by considering a linear decay in shoreline tilting and water levels taken from the shore displacement curve in Figure 8.
Results and interpretations
Sediment stratigraphy and chronology of the coastal landforms and ship burials in Salme
Sediments of the Salme coastal landforms, in which two Vendel Period ship burials have been found, were studied in three N-S and W-E orientated sections as presented in Figures 1c and 2. Two beach deposit units can be distinguished on top of the bluish-grey varved clay by GPR and coring data. The lower, 3–5-m-thick unit consists of beige sandy-gravelly sediment, which contains unbroken mollusc shells of Macoma baltica, Cerastoderma glaucum and Lymnaea stagnalis. A fragment of a Cerastoderma glaucum shell from the beach deposits in trench OSL3 at an elevation of 1.45 m a.s.l. was radiocarbon dated to between 3490 and 3330 BC. This age is further supported by the OSL age of 3320 ± 370 BC which was obtained from the same trench and the same sedimentary unit (Table 1). Sediment stratigraphy, GPR and LiDAR data suggest that these sediments were accumulated in connection with the S-N-orientated spit formation to the north of the Salme Strait. The spit, with an axis which today follows the topographic high, was initially relatively narrow but became progressively wider by growing westwards (the eastern side of the spit is being eroded by the present coastal action of the sea). GPR sections show a generally westerly-tilted layering with some stronger reflectors which mark out temporary seabed stands (Figure 2a). Reflectivity within the lower unit appears to be higher close to the axis of the spit, indicating higher energy deposits, whereas farther from the spit axis the apparent tilt angles become gentler and reflectivity weaker, suggesting deposition in calmer conditions. On both the northern and southern sides of the strait the reflectors are steeply tilted towards the channel as the upper surface of the varved clay deepens.
The upper unit also consists of sandy-gravelly beach deposit which contains unbroken mollusc shells and is separated from the lower unit by a wavy reflector on the GPR profile (Figure 2). While the lower unit shows tilted layering and deposition in deeper water, the upper unit shows deposition in the shallow sea near the coastline. This is also supported by coring data at the GPR profile ‘A’, where the upper unit can be distinguished from the lower unit by coarser sediments which refer to sedimentation in higher energy conditions. Two OSL dates from this unit suggest the formation of the upper part of the landform around 60–320 AD (Table 1), which is about 710–450 years older than the ship burials. However, a piece of charcoal in the beach deposits, which was dated to around 200 AD, also falls into the time window of the final formation of the coastal landform at Salme. A thin sandy-gravelly topsoil layer with a thickness of ca. 0.1–0.4 m has developed on top of the beach deposits.
The Salme palaeostrait
Coastal terraces on the shores of the Salme Strait
In LiDAR topography, two main coastal terraces can be identified bordering the Salme palaeostrait (Figure 10c). The foot of the upper terrace was mapped out at an elevation of ca. 5.4 (or between 4.4 and 6.3) m a.s.l. and the foot of the lower terrace at an elevation of ca. 3.0 (or between 2.3 and 3.9) m a.s.l. (Table 2). On the northern coast, which is more exposed to wave action, the terraces are about 0.2–0.1 m higher when compared to the south-eastern and north-eastern coasts.
Sediment stratigraphy and the chronology of the Salme Strait deposits
According to the coring data, sandy deposits dominate in the open and wider western part of the Salme palaeostrait whereas in the narrow eastern part fine-grained organic-rich sediments (laminated silty gyttja) also occur in a relatively limited area (Figure 3a).
(a) Geological profiles A-B and C-D with coring sites and trenches (Figure 1c) and (b) sediment stratigraphy of the master core SV02 from the infill of the Salme Strait, with the age-depth model based on AMS radiocarbon dates, LOI and grain-size results.
A 2.5-m-long master core was taken from the left bank of the River Salme which comprised four sedimentary units (Figure 3b). Sediments of the lowermost unit (unit ‘A’) consist of a 2-cm-thick gravelly sand layer with shell detritus laying on top of the glaciolacustrine varved clay. This thin sand layer is covered by a 150-cm-thick layer of laminated silty-sandy gyttja (unit ‘B’). Organic content in the sediments of unit ‘B’ typically remain below 10%, and at around 7% on average. The mineral component of the layered sediments of unit ‘B’ is very fine (silty), poorly sorted sand in the upper 0.8 m interval, and fine to medium, poorly or moderately sorted sand below it. Two uppermost units are sandy organic-rich floodplain sediment layers, separated according to organic content to lower sandy peat with an organic content of 12–29% (unit ‘C’), and to upper sandy peat with an organic content of 42–67% (unit ‘D’). The thickness of both units is 15 cm.
Altogether, eight samples of the terrestrial seeds and charcoal from unit ‘B’ were AMS radiocarbon dated (Table 3, Supplemental Material, available online). Five of these samples were used to construct an age-depth model of the master core (Figure 3b), which is useful, as it allows to estimate the age of the sections that are not directly dated, such as, for example, the ages of sediment boundaries. Two lowermost samples from the depth of 150–200 cm, which showed ages which were too recent when compared to the main set of AMS samples, and one sample at a depth of 45 cm which showed too old age, were rejected from the modelling. According to the age-depth model, the accumulation of the layered sediments of unit ‘B’ lasted for about 360 years, between 920 and 1280 AD (Figure 3b), and therefore accumulation started about 170 years after the burial of the Salme ships.
Diatom stratigraphy
Altogether 12 diatom samples were analysed from the Salme master core. A total of 132 diatom taxa belonging to 55 genera were identified in the 188-cm-long sediment sequence which covers a time span of about 380 years (920–1300 AD). The relative frequency of diatoms with an abundance of at least 2% or more is presented in Figure 4. In general, the diatom composition of the laminated silty gyttja of unit ‘B’ is characteristic of diatoms which are found today in the littoral zone of the Baltic Sea, rich with macrophytes (Vilbaste et al., 2000; Witkowski, 1994; Witkowski et al., 2000), and hence represents a shallow, marine or brackish water environment. The diatom stratigraphy was divided into three diatom assemblage zones (DAZ), defined by cluster analysis (CONISS). The DAZ-I was divided into two sub-zones (DAZ-Ia and DAZ-Ib) in order to highlight a different environment at the lowermost part of the DAZ-I in comparison to the rest of the zone, as indicated by a rapid shift in dominant diatom taxa.
DAZ-I (208–100 cm; 920–1140 AD): the lowermost sample (208 cm; 920 AD) in DAZ-Ia contains a significant proportion of marine/brackish/halophilous taxa (48%), and indifferent taxa (28%). The most abundant marine- and brackish-water epiphytic diatoms, such as Tabularia tabulata, T. fasciculata, Rhoicosphenia abbreviata, Cocconeis scutellum, Grammatophora oceanica, G. marina, Licomorpha spp., and indifferent epiphytic Cocconeis pediculus and C. placentula var. euglypta are found on the macrophytes (e.g. Pilayella littoralis, Ruppia maritime, Ceramium sp., Cladophora sp.) growing in the coastal areas (Witkowski, 1994; Witkowski et al., 2000). The amount of marine/brackish diatoms in the rest of the samples of the DAZ-Ib decreases (25–32%), but small fragilarioid taxa with brackish-water affinity (e.g. Opephora mutabilis, O. krumbeinii, Fragilaria sopotensis) increases (ca. 36–44%). In general, diatom composition indicates a brackish coastal area with gradually decreasing macrophyte stands. Moreover, an increase in periphytic diatoms is observed at 110 cm (1110 AD). This may have been caused by physical disturbances (of either natural or anthropogenic origin) along the coastal area.
DAZ-II (100–30 cm; 1140–1270 AD): DAZ-II is dominated by small fragilarioid taxa with a brackish-water affinity which reaches up to a maximum of 55%, although marine/brackish (13–27%) and indifferent (12–16%) taxa still form a significant proportion of the diatom assemblage. The highest concentration of diatom valves is detected at a depth of 90 cm (ca. 1150 AD). In the upper part of the DAZ-II, marine/brackish taxa gradually decrease (down to 9%), but small fragilarioid taxa which favour a freshwater environment increase (up to 24%). Changes in diatom assemblages from brackish to freshwater loving ones point to a decrease in marine influence.
DAZ-III (30–20 cm; 1270–1300 AD): DAZ-III is dominated by very small fragilarioid taxa which prefer a freshwater environment (Staurosira venter 15–27%, Staurosirella mutabilis 19–22%). In this zone most of the benthic diatoms are heavily eroded, and physically broken into pieces. A mass occurrence of small fragilarioid taxa, which are pioneer species with the ability to successfully survive in a rapidly changing environment (Yu et al., 2004), coupled with a decline in marine/brackish periphytic diatoms, is considered to be an indicator for a reduction in marine/brackish conditions (Grudzinska et al., 2012, 2013), potentially resulting from a fall in relative sea level. Also, changes in lithology (thin alternating silt and gyttja layers which are replaced by organic-rich silt and sand) in the transition from DAZ-II to DAZ-III is a sign of the gradual closure of the strait.
Summary diagram of selected diatom taxa from Salme master core (SV02) with modelled ages. Diatoms are arranged according to their salinity preference. Diatom assemblage zones (DAZ) are defined by the results of cluster analysis (CONISS).
SEM analysis of the sand grains
SEM analysis of the master core sediments of units ‘B’–‘D’ reveal three types of quartz sand grains in the investigated samples, in the form of angular, rounded, and subangular grains (Figure 5a–c, respectively). These are seen throughout most of the investigated samples except for those samples which come from a depth of 130and 140 cm, where rounded and angular grains respectively are absent (Figure 6).
Explanatory images of quartz sand grain shapes and microtextures: (a) angular grain, (b) rounded grain, (c) subangular grain, (d) grain with big flat cleavage surface, (e and f) big conchoidal features with straight and arcuate steps, (g) more intense occurrence of V-shaped percussion marks, (h) weaker occurrence of V-shaped percussion marks on grain surface, (i) dulled surface, (j and k) oriented etch-pits on different crystal planes and (l) precipitation in grain depression.
The occurrence of selected quartz grain microtextures and SEM-based sediment origin with sample labels and their positions in the simplified SV02 master core profile.
In general, subangular grains dominate in the investigated samples except for the bottommost 200 cm sample, where a combination of angular and rounded grains occurs (Figure 6). Among microtextures which are of mechanical origin, flat cleavage surfaces are seen on grain surfaces (Figure 5d), along with large conchoidal features (Figure 5e and f) and straight and arcuate steps (Figures 5e, f and 6). The V-shaped percussion marks are features that can commonly be found on grains (Figure 5g). However, their frequency may differ (Figures 5h and 6). A chemically-induced dulled surface (Figure 5i) is seen in all samples, but the occurrence of orientated etch pits seem more variable. Similarly, precipitation occurrence varies (Figure 5), and can be traced not only at the top of the grain surface, but also in grain depressions (Figure 5j).
Numerous microtextures are detected on investigated quartz grains, and their identification is used to track a sedimentary environment across a wide spectrum of its types (for details see Table 2 in Vos et al., 2014). Practically only four types of microtextures help to understand marine deposition conditions and an intense wave action: fresh surfaces, V-shaped percussion marks, adhering particles and dissolution (Costa et al., 2012). Therefore, those quartz grains which have been reworked by storm action carry more percussion marks and fresh surfaces than do their potential source material (Bruzzi and Prone, 2016; Costa et al., 2017), whereas orientated etch pits constitute a group of chemical microtextures which result from alkaline fluids and seawater (Bull, 1981). These important marine markers have been seen on grains from both ancient and recent coastal sediments around the Baltic Sea (Kalińska-Nartiša et al., 2017, 2018a, 2018b).
Following this knowledge, part of the profile between 40 and 150 cm and additionally at its bottom at 200 cm indicate marine conditions being, however, changeable, and therefore marking alternating periods of stronger (stormy; Figure 7a) and weaker sea action. This is mostly seen through a generally high number of orientated etch pits, which seriously alternate among the samples. Two samples (40 and 90 cm) are particularly of marine origin, with the highest number of etch pits, and therefore may mark a period of saline water intrusion (Figure 7b). Additionally, a sample from a depth of 90 cm may mark increased storm action if one is following instructions by Costa et al. (2012), since the number of fresh and flat cleavages and cracked surfaces atop quartz grains dominate and this is also similar to a sample from 200 cm (Figure 7a).
Explanatory images of quartz sand grain microtextures as typical for the Salme profile: (a) grains of likely storm origin with sharp edges, fresh microtextures and high grain relief, (b) grains from a highenergetic marine environment with abundant occurrence of oriented etch pits, (c) fluvial grains with dulled surfaces and pits and (d) grains that underwent post-depositional weathering, so their edges are not fresh anymore.
Two samples in the profile, from depths of 30 and 190 cm, clearly stand out in the investigated set. This is seen through the smallest number of marine-origin orientated etch pits on the grain surface. Similarly, the beach-ridge sample carries the same property, as can be seen through diminishing marine impact upon grain transportation (Figure 7c), thereby displaying a fluvial record rather than a marine or storm one (Figure 6). Importantly, part of the microtextures which are of mechanical origin and which are normally fresh (as resulting from intense wave action), seem old and weathered by chemical processes which took place following the mechanical breakage, as seen in the samples at 10 and 130 cm (Figure 7d).
Grain-size results show few distinct peaks of fine sand fraction at depths of 110, 150 and 200 cm, which are in contrast, for example, with the depth of 130 cm where the share of fine sand is smaller (Figure 3b). A combination of grain-size and SEM studies show that finer grains tend to be more angular (Fuller and Murray, 2002; Mahaney, 2002; Molen, 1992) and this also seems true for our results. Two samples, where a finer fraction dominates, carry more grains with a high surface relief, which is something that is likely linked with an increased number of angular grains.
A large part of the investigated profile coincides chronologically with the most widespread stormy periods during the Holocene, and especially with the Holocene storm periods (HSP) IV and V (Figure 8) which occurred in northern coastal Europe at 50–900 AD and 1350–1700 AD, respectively (Sorrel et al., 2012). Assuming that two investigated samples reveal a stormy origin, two storm periods took place in the eastern Baltic at about 920 AD and 1150 AD, respectively. While the first record corresponds well with the HSP IV which was also detected around the neighbouring island of Hiiumaa ca. 1200 years ago (Vilumaa et al., 2016), the second is not really seen either in the Baltic palaeostorm record or in that of the Atlantic Ocean (see for details Pouzet et al., 2018) except for the period between 1150 and1650 AD, when major storminess occurred in Britain (Orme et al., 2016). The most recent study from the southern Baltic Sea shows a lack of storm deposits along the coast up until to AD 1500 when the first storm deposits occurred in the Gulf of Gdańsk (Moskalewicz et al., 2020) and in the Puck Lagoon (Uścinowicz et al., 2020), meaning that our data set adds another chronological puzzle to the Baltic Sea palaeostorm record prior to the Little Ice Age.
(a) Late-Holocene RSL curve for Salme (blue line) compared with the curve of the Pomeranian coast (Lampe and Janke, 2004) and with the absolute land uplift rate. Late-Holocene storm periods III, IV and V according to Sorrel et al. (2012) are marked as grey zones and the alternation of warm and cold periods according to Reimann et al. (2011) and Neukom et al. (2019) are marked by blue and red zones at the top of the graph (LIA – Little Ice Age, DACP – Dark Age Cold Period, MWP – Mediaeval Warm Period, RWP – Roman Warm Period) and (b) variations in the annual average relative sea levels in Virtsu (western Estonia) (modified from Tõnisson et al., 2019).
Discussion
Late-Holocene RSL changes around Saaremaa Island
Results from Salme and the neighbouring archaeological and geological sites (Figure 1b) were used to reconstruct Late-Holocene RSL changes for Saaremaa Island (Figure 8). Altogether, 33 sea-level indicators were analysed (Table 3), including some previously published data from Salme burial sites (Konsa et al., 2009; Peets and Maldre, 2010), from the nearby Vintri amber deposit (Nirgi et al., 2017; Ots, 2012), from the fortified settlements in Asva (Lang and Kriiska, 2001) and Kuressaare (Püüa, 2013), and a typological age determination from the Viltina harbour site (Mägi, 2009). Annual average RSLs from the Virtsu tide-gauge station (Figures 1b and 8b) for the 20th century were also used in the RSL reconstruction (Tõnisson et al., 2019). Vertical and chronological uncertainties in the RSL data were re-evaluated using the HOLSEA database format (Hijma et al., 2015; Shennan et al., 2018) and are presented in the Supplemental Material, available online. Vertical uncertainties in the geological RSL data are around ±0.3 m and are related to sampling uncertainties and elevation (LiDAR) data uncertainties. Vertical uncertainties in the archaeological data, depending upon the size and topography of the individual site, remain between ±0.30 and ±1.04 m. Together with the indicative range of the uncertainty level (Hijma et al., 2015), the total vertical uncertainty was estimated to be around ±0.5 m for geological RSL data, and ±1.0 m for archaeological RSL data. Three radiocarbon dates from the Salme strait and one from the burial site were rejected due to uncertainties in their chronology (Table 3; Supplemental Tables, available online). Two dates of gaming pieces were also set aside because they do not reflect the age of the burial, and two dates from Salme coastal sediments are older than the main time-range of this study (Table 3). Eventually, a total of 25 RSL indicators, including seven RSL index points, 13 terrestrial limiting points and five marine limiting points were used to reconstruct the RSL curve which is shown in Figure 8.
The RSL curve shows a nearly linear regressive trend since the Bronze Age due to the continuous glacial isostatic adjustment outpacing the declining rates of post-glacial absolute sea-level rise. Two radiocarbon dates from the buried organic bed in coastal deposits from the Vintri site (Nirgi et al., 2017) show the RSL to be around 5.5 m a.s.l. at about 700 BC. New OSL ages from the upper part of the Salme coastal deposits suggest an RSL fall to an elevation of ca. 4 m a.s.l. during the following 700 years. Elevations of the Viking Age trading site at Viltina (1000–1150 AD) with remains of the jetty construction (Mägi, 2009) suggest a further 2–2.5 m lowering in the RSL during the next 900 years.
New chronological and diatom data from the Salme master core show an initial reduction in the marine/brackish environment at 920–940 AD in the Salme Strait. The following gradual increase in freshwater diatom species suggests the beginning of the strait’s closure. As evidenced from the freshwater floodplain deposits (DAZ-III) in the master core, the strait was closing, and the River Salme started to flow through the valley sometime after 1270 AD. Considering these changes, it is likely that the RSL dropped below the threshold elevation at 0.75 m a.s.l. (the upper boundary of the unit ‘C’) at around 1300 AD.
RSL at an altitude of ca. 0.8 m in the Salme area around 1300 AD is also supported by the earliest boardwalk constructions in a neighbouring Kuressaare fortified settlement, which are dated back to 1320–1350 AD and which help to suggest that the RSL was less than 1 m a.s.l. at that time. Dated boardwalk constructions were initially established slightly above the mean sea level but were then damaged several times, most likely during storms which were accompanied by higher water levels, as evidenced from archaeological excavations (Püüa, 2013).
Overall, the average RSL fall during the period since 700 BC was about 2 mm/year, being ca. 0.3 mm/year slower when compared to the present-day absolute land uplift rate. This difference is most probably attributed to slow eustatic sea-level rise during the Late-Holocene (Lambeck et al., 2014). The Late-Holocene RSL curve from the Pomeranian non-uplifting coast (Lampe and Janke, 2004) suggests that sea-level has risen in the Baltic Sea by about 90 cm since 700 BC (Figure 8). If this value is to be implemented for the Saaremaa RSL curve, the total uplift for the area would amount to 6.2 m, with an average uplift rate of 2.3 mm/year (Figure 8), which is the same figure that is derived from present-day geodetic observations for Saaremaa Island (Vestøl et al., 2019). Therefore, it can be concluded that postglacial land uplift since 700 BC has been somewhat stable, and with some fluctuations in the Late-Holocene RSL probably being related to climatic events including cyclic changes in the strength of the westerly atmospheric circulation (Suursaar et al., 2006). Our Late-Holocene RSL data for Saaremaa is still rather scarce when it comes to relating RSL changes to climatic events. Still, it seems that some slowdown in regression occurs somewhat after 1300 AD, which may be attributed to accelerated sea-level rise after the Little Ice Age and during the industrial period. This is also consistent with the Virtsu tide-gauge data from the 20th century which reveals a near-zero RSL trend (Figure 8; Tõnisson et al., 2019).
The palaeogeography of the Salme area and the geoarchaeology of the Salme ship burials
Palaeogeographic shoreline reconstructions based on RSL change for the Salme area (Figures 9 and 10) show the development of the former Salme Strait, which separated the Sõrve Paleoisland from the Saaremaa mainland, between 700 BC and 1300 AD. In the Late Bronze Age (Figure 10a), at about 700 BC, the strait was up to 2 km wide and ca. 5 m deep and was probably part of the Bronze Age transportation network (Nirgi et al., 2017). In the south it was bordered by the Läätsa palaeoisland, and from the east by the slowly growing south-north orientated spit system.
Palaeogeographic reconstruction of Saaremaa Island with archaeological sites, and indication of RSL isobases (m a.s.l.) at around 700 AD.
Palaeogeographic reconstructions showing the development of the Salme Strait: (a) from the Bronze Age at 700 BC, (b) from the Vendel Period at 750 AD, when the ships had been buried to Salme and (c) from the period between 1100 and 1300 AD, before the isolation of the Salme Strait. The white areas are marking the present sea.
A sediment stratigraphy and the GPR data of the spit displays two units of coastal deposits (Figure 2). The properties of the lower unit indicate deposition in deeper, calmer water, when compared to the upper unit. The westward tilting of the layering with gentler tilting further to the west from the spit axes points to the longshore sediment transport in the east due to wave refraction around the Sõrve and Läätsa palaeoislands in the south (Figure 10a). Longshore sediment transport from the south is supported by the fact that the source of sandy gravelly material is available only from the south of Salme area, and neither are similar deposits available along the bottom deposits east of Salme (Tõnisson et al., 2007). In addition, the present-day S-N longshore drift has continuously been monitored in the area (Orviku et al., 2009; Tõnisson et al., 2007). According to our OSL and radiocarbon dates, the formation of this underwater bar started as early as the 3300–3500 BC period (Figure 2). The upper unit accumulated on top of this initial landform in shallow water with a high energy level of wave activity (Figure 2). Such a sedimentary environment favoured the formation of a series of SE-NW orientated spit appendixes to the north of the palaeostrait (Figure 10), pointing to longshore transportation from the south along the eastern side of a spit. The distal, classical arched ends of these appendixes are turned towards the NE, which can only be explained by storm events which were directed SW-NE, and which reached the palaeostrait from its more open and wider SW side. Similar features to the south of the Salme palaeostrait are not visible in the current topography, which can be explained by the slightly higher topography (1.5–2 m; Läätsa palaeoislet) at the final stage of the spit formation when compared to the northern coast of the strait. This all leads us to believe that the narrowing of the strait to the east was not due to classical spit development which resulted from longshore transport but was instead due to the initial landform already being present, with the closing of the strait taking place due to uneven rebound and the water-level lowering. As derived from the OSL dates, the final stage of spit formation to the north took place between 60 and 320 AD (Figure 2) and, in concert with the rebound, it caused a slow closing of the strait about 1270–1300 AD.
At the time of the presumed battle and the following ship burials at about 650–770 AD, a semi-enclosed strait existed around Salme with a wider western section and a narrow eastern section (Figures 9 and 10b). The western part of the strait had a sandy bottom. It was about 1 km wide and had a water depth of more than 2 m. The wind protected eastern part of the strait was about 80–100 m wide, with a sandy and gravelly bottom and with water depths of up to 2.8 m. The relatively steep and wind-protected sandy and gravelly shores in the narrow part of the strait were probably the best places in the Salme area for landing the Viking ships (Figure 10b). The narrowing of the Salme palaeostrait occurred about 1270–1300 AD (Figure 10c), which can be supported by sedimentological evidence, as well as by the diatom data. The mass occurrence of diatom species with the ability to survive successfully in a rapidly changing environment, coupled with the decline in marine/brackish periphytic diatoms is considered an indicator of the isolation (Grudzinska et al., 2012, 2013). Furthermore, the SEM results for quartz mineral grains from a depth of 30 cm stand out in terms of the smallest number of marine-origin orientated etch pits on the grain surface, indicating a clearly diminishing marine impact on grain transportation and rather arguing for a non-marine influence such as, for example, one which emerged from the fluvial and somehow less-energetic environment (Figure 7c).
Our RSL and palaeogeographic reconstructions show that the Salme ship burials were not initiated in the submerged section of the Salme Strait, and nor were they initiated along the strait’s coastline, or even along the eastern coast of the Salme spit, as was discussed following archaeological excavations. Our reconstructions show that the burials are located about 2–2.5 m above coeval sea level (Figure 8) which makes more than 100 m from both, the shore of the former strait in the south, as well as from the sandy-gravelly palaeocoast of the Baltic Sea in the east (Figure 10b).
The Salme I and II ship burials were carried out in sandy-gravelly coastal deposits (Figure 2) which accumulated along the open coastal zone at about 710–450 years earlier than the burial event. Therefore, it is likely that both ships were moved from the shore to the higher ground for the burial. The orientation of the vessel hulls suggests that they were most likely moved to the burial place from the shore of the narrow Salme Strait rather than from the coast of the Baltic Sea to the east (Figure 10b).
Conclusions
− A Late-Holocene RSL curve was constructed for Saaremaa Island, based on geological and archaeological proxies, and covering the last ca. 3000 years. The curve shows an almost linear RSL fall from 5.5 to 0.8 m a.s.l. between 1000 BC and 1300 AD, at an average rate of 2 mm a year.
− The postglacial land uplift since 700 BC has been somewhat stable and even, with only some minor fluctuations in the Late-Holocene RSL which were probably related to climatic events which included cyclic changes in the strength of the westerly atmospheric circulation.
− A slowdown in regression occurs somewhat after 1300 AD, which may be attributed to accelerated sea-level rise after the Little Ice Age, and which continues during the industrial period, being consistent with the tide gauge measurements from the 20th century.
− Palaeogeographic reconstructions show that, during the burial of the Salme ships at around 650–770 AD a semi-enclosed strait existed in Salme with a wider western and a narrower eastern section. The wind-protected eastern part of the strait was about 80–100 m wide, with a sandy-gravelly bottom and a water depth of up to 2.8 m. The relatively steep shores in the narrow part of the strait were probably the best places in the Salme area for landing Pre-Viking Age ships.
− The isolation of the Salme palaeostrait occurred about 1270–1300 AD, which can be supported by sedimentological evidence and the diatom data.
− The Salme I and II vessels were buried in sandy-gravelly coastal deposits which had accumulated in the open coastal zone about 710–450 years before the burial event.
− The major archaeological significance of the study comes from the conclusion that, when considering all of the palaeogeographical evidence and the coastline reconstructions, the burials of the Salme I and Salme II vessels did not take place through a process of them being submerged; instead, they were probably located about 2–2.5 m above the coeval sea-level and more than 100 m from the strait’s shore. Therefore, it is likely that both ships were moved from the shore of the gradually-closing strait to the higher ground for the burial.
Supplemental Material
sj-xlsx-1-hol-10.1177_09596836211066596 – Supplemental material for Late-Holocene relative sea-level changes and palaeoenvironment of the Pre-Viking Age ship burials in Salme, Saaremaa Island, eastern Baltic Sea
Supplemental material, sj-xlsx-1-hol-10.1177_09596836211066596 for Late-Holocene relative sea-level changes and palaeoenvironment of the Pre-Viking Age ship burials in Salme, Saaremaa Island, eastern Baltic Sea by Triine Nirgi, Ieva Grudzinska, Edyta Kalińska, Marge Konsa, Argo Jõeleht, Helena Alexanderson, Tiit Hang and Alar Rosentau in The Holocene
Supplemental Material
sj-xlsx-2-hol-10.1177_09596836211066596 – Supplemental material for Late-Holocene relative sea-level changes and palaeoenvironment of the Pre-Viking Age ship burials in Salme, Saaremaa Island, eastern Baltic Sea
Supplemental material, sj-xlsx-2-hol-10.1177_09596836211066596 for Late-Holocene relative sea-level changes and palaeoenvironment of the Pre-Viking Age ship burials in Salme, Saaremaa Island, eastern Baltic Sea by Triine Nirgi, Ieva Grudzinska, Edyta Kalińska, Marge Konsa, Argo Jõeleht, Helena Alexanderson, Tiit Hang and Alar Rosentau in The Holocene
Footnotes
Acknowledgements
The authors are grateful to Hando-Laur Habicht for his invaluable assistance at fieldwork, to Leeli Amon-Veskimeister for helping to sort out terrestrial plant macros for radiocarbon dating and to Garel Püüa for introducing the archaeological objects in Kuressaare. They also thank Pedro JM Costa and an anonymous reviewer for their comments and suggestions for improving the manuscript.
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: Studies were financed by Swedish Research Council via MHVAJ18304 ‘Studies on the boat burial of Salme I and its paleoenvironment’, which is part of the ‘Viking Phenomenon’ project (Uppsala University), and supported by the Estonian Research Council grant PRG1471. This paper was written in the framework of IGCP project 639 ‘Sea Level Change from Minutes to Millennia’, supported by UNESCO and IUGS.
Supplemental material
Supplemental material for this article is available online.
References
Supplementary Material
Please find the following supplemental material available below.
For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.
For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.
