Halmyris: Geoarchaeology of a fluvial harbour on the Danube Delta (Dobrogea,Romania)

Abstract

In Northern Dobrogea, north of the Dunavăţ promontory, the Roman fortress of Halmyris was founded in the late 1st century AD on a Getic settlement dating to the middle of the 1st millennium BC, probably associated with a Greek emporium of the Classical and Hellenistic periods. At the time of the foundation of Halmyris, the Danube delta had already prograded several kilometres to the east leading to the progressive retreat of the sea and the formation of a deltaic plain characterised by numerous lakes and river channels. Here, we present the results of a multiproxy study combining sedimentology and palaeoecology to (1) understand the evolution of fluvial landscapes around Halmyris since ca. 8000 years BP and (2) identify the fluvial palaeoenvironments close to the city in Getic/Greek and Roman times, in order to locate and characterise the waterfront and the harbour. Our overriding objective was to improve understanding of human–environment relations in river delta settings. We demonstrate that Halmyris, protected by the Danubian floods due to its location on a palaeo-cliff top, had direct access to the river. A secondary channel of the Saint George, flowing north of the site, has been elucidated between the 7th century BC and the 7th century AD and could have been used as a natural harbour.

Keywords

Black Sea Danube delta fluvial harbour geoarchaeology geomorphology Halmyris

Introduction

In recent decades, deltaic environments have attracted interdisciplinary research interest, looking to understand human–environment interactions in these important waterfront areas (see Anthony et al., 2014). These lowlands at the land–sea interface were particularly attractive, since the stabilisation of sea-level around 6000 yr BP, because they provided fertile lands for agriculture and a permanent freshwater supply, as well as access to the hinterland by fluvial routes. A multi-proxy approach allows to clearly identify the environmental pressures that affected deltaic harbour cities and to highlight the adaptation of populations, including, for instance, Alexandria (Nile delta: Flaux et al., 2017; Goiran et al., 2005), Miletos (Büyük Menderes delta: Brückner et al., 2006, 2014), Ephesos (Küçük Menderes delta: Stock et al., 2013, 2016), Ainos (Maritsa delta: Seeliger et al., 2018), Pisa (Arno and Serchio delta: Allinne et al., 2016; Benvenuti et al., 2006; Kaniewski et al., 2018), the Greek colonies of the Taman Peninsula (Kuban delta: Giaime et al., 2016; Kelterbaum et al., 2011); Aigues-Mortes (Rhône delta: Rey et al., 2016), Ostia (Tiber delta: Goiran et al., 2014; Salomon et al., 2017, 2018) or Utrecht (Rhine delta: Van Dinter et al., 2017).

The Danube delta is one of the largest fluvial and wave-dominated delta’s in the world and constitutes one of Europe’s major wetland zones. It has been listed on the UNESCO World Heritage List since 1991. Occupied since the Neolithic (Carozza et al., 2012, 2013; Micu et al., 2009), the coast of the Danube delta has a long and rich history of human occupation. It was, at the end of the Archaic Period, coveted by the Black Sea’s first Mediterranean colonists. The area offered favourable environmental conditions for thriving socio-economic activities (Avram et al., 2004). In recent years, geoarchaeological studies looking to understand environmental changes at several ancient sites located on the southern margin of the Danube delta have been undertaken (e.g. Orgame, (Bony et al., 2013, 2015; Histria, (Bivolaru et al., in press; Vespremeanu-Stroe et al., 2013; Enisala and Babadag, (Preoteasa et al., in press; see Figure 1a for location). These different studies have highlighted the interplay between environmental and anthropogenic changes, furnishing new data on the evolution of the ancient sites. In contrast, and despite the geoarchaeological importance of the Lower Danube, its ancient and medieval fluvial harbours are poorly known.

Figure 1.

(a) Location of Halmyris in the centre of the Danube delta and of other archaeological sites cited in the text. The sand spit shows the limit between the fluvial and the marine delta. Age of the lobes from Giosan et al. (2006) and Vespremeanu-Stroe et al. (2017). Image ESRI/CNES. Insert from Revenga et al. (1998) showing the Danube watershed. (b) Reconstruction of the progradation of the Old Danube lobe into the Danube Bay and of the initial spit formation (age in yrs cal. BC). Topography from Landsat image, 2013 (From Vespremeanu-Stroe et al., 2017).

River delta sedimentary archives are particularly interesting because they can help to probe the interplay between river course evolution (fluvial geomorphology) and archaeological development. In this article, we present a case study to highlight these human–fluvial landscape interactions during Antiquity. We focus on the archaeological site of Halmyris, located ca. 38 km from the present mouth of the Saint George arm of the Danube, on the northern side of the Dunavăţ promontory (Figure 1a). Several authors have suggested the presence of a harbour connected to the Danube River directly north of the fortress in Roman times (Bounegru and Zahariade, 1996; Suceveanu et al., 2003; Zahariade, 1991). By coring on the delta plain, 100 m from the main northern gate of the fortress, we looked to better understand the evolution of the environment during the last 8000 years. Our investigations are based on multiproxy analyses of fluvial sedimentary archives (sedimentology, molluscs, ostracods and chironomids). The combination of such proxies is particularly useful in palaeohydrology (Petts et al., 1989). Ostracods and chironomids have proven to be helpful proxies in the study of palaeoenvironments in hypohaline river-delta settings (e.g. to reconstruct the past connectivity between alluvial waterbodies and the main channel; to estimate shifts in salinity; Gandouin et al., 2005, 2006). Our aim was to study the stratigraphic sequence in order: (1) to understand the long-term evolution of the sedimentary environments and (2) to elucidate the presence of a harbour sedimentary sequence corresponding to the occupation of the site by the Getic and Greeks (6th to 1st centuries BC) and the Romans (1st to 7th centuries AD).

Geomorphological context

The Danube river is the most important water and sediment supplier of the Black Sea, with a water discharge of 190 km³/year and a sediment discharge of 25–35 million tonnes per year (4–6 million comprising sands) (Panin et al., 2016; Panin and Jipa, 2002). The delta constitutes a large coastal accumulation of sediments that has evolved continuously during the Holocene, under the combined effects of natural and human-induced pressures. The river started to build its delta in the Danube Bay around 8000–7500 cal. yr BP (6000–5500 cal. yr BC; Vespremeanu-Stroe et al., 2017). The deltaic sediments fill the pre-Dobrogea depression that lies mainly on the Scythian platform (Burchfiel et al., 1974). This important accumulation of sediments (also supplied by important loess formations; (Fitzsimmons and Hambach, 2014; Marković et al., 2015) has led to subsidence. Estimates for subsidence rates vary between 0.4 and 0.6 mm yr⁻¹ (Vespremeanu-Stroe et al., 2017) and 1.3–2 mm yr⁻¹ (Giosan et al., 1997).

According to Antipa (1914), Panin and Jipa (2002), Stefanescu (1982), and Vespremeanu-Stroe et al. (2017), the delta can be divided into two distinct geomorphological units: (1) the western fluvial delta and (2) the eastern and southern marine delta. In their recent study, Vespremeanu-Stroe et al. (2017) confirm that the limit between the two units is formed by a coastal spit that prograded southwards in the northern part of the delta after ca. 6700–6100 cal. yr BP (4700–4100 cal. yr BC; Figure 1a and b). The formation of the fluvial delta can be divided into two main phases. Initially, north of the Dunavăţ promontory, the old Danube delta (mainly formed by the Saint George arm) started to build its first lobes around 8000–7500 cal. yr BP (6000–5500 cal. yr BC). Reconstruction of the palaeo-delta front position shows that the area, north of ancient Halmyris, was washed by the sea until almost 7500 cal. yr BP (5500 cal. yr BC; Vespremeanu-Stroe et al., 2017; Figure 1b). The important sedimentary input between 7500 and 6500 cal. yr BP (5000–4500 cal. yr BC) gradually led to the progradation of the delta front (1.5–2 m yr⁻¹; Vespremeanu-Stroe et al., 2017). Formed from 6700 to 5800 cal. yr BP (4700–3800 cal. yr BC), the spit did not influence the early stages of delta formation. In the mid-Holocene, the Danube delta was mainly formed by the Saint George arm, the oldest arm of the Danube and which has been continuously active for more than 8000 years. The second major phase in the evolution of the fluvial delta started around 5500 cal. yr BP (3500 cal. yr BC) and was characterised by fluvial and peat aggradation in the shallow lakes spanning a large part of the amphibic delta plain. This transformation took place in a context of low relative sea-level rise (ca. 1 m between 5500 and 3500 cal. yr BP (3500–1500 cal. yr BC); Vespremeanu-Stroe et al., 2013) and low subsidence (0.4–0.6 mm yr⁻¹; Vespremeanu-Stroe et al., 2017).

The formation of the ‘marine delta’ started eastward and southward of the initial spit/barrier and is composed of five main open coast lobes that have contributed to the present coastal geography of the delta (Giosan et al., 2006; Panin, 2003; Stefanescu, 1982; Vespremeanu-Stroe et al., 2017). The first phase of this ‘marine delta’ was initiated by the first open coast lobe formation (Old Saint George; Figure 1a) between 6500 and 5500 cal. yr BP (4500–3500 cal. yr BC), characterised by increasing wave action on the coast (Vespremeanu-Stroe et al., 2017). The formation of the different deltaic lobes had no direct influence on the site of Halmyris because of its location on the old Danube delta plain (‘fluvial delta’). In fact, the delta front has been situated downstream of Halmyris for around 6000 years. Nonetheless, the harbour environments of other ancient settlements, including those located in the southern part of the delta (Razelm-Sinoe lagoon), were directly impacted by the progradation of the various branches of the river and their erosion/reworking induced by longshore drift (Bony et al., 2015; Preoteasa et al., in press; Vespremeanu-Stroe et al., 2013).

Historical and archaeological contexts of Halmyris

The ancient history of Halmyris is divided into three main occupation phases (Suceveanu et al., 2003; Zahariade et al., 1987; Zahariade and Phelps, 2002).

Between the 6th and the 1st centuries BC, the site was occupied by a Getic settlement. During this period, archaeologists have highlighted the presence of a possible Greek emporium associated with this settlement. The emporium was integrated into the chora of Histria, or, more likely into that of Orgame (Suceveanu et al., 2003; Zahariade, 1991; Zahariade and Phelps, 2002). The authors base their hypothesis on the toponymy (Halmyris is possibly a Greek name, related to the ancient homonymic gulf, which could mean salt water) and on the Greek pottery discovered (especially amphorae from Chios, Chersonessos and Thassos). Even though the hypothesis of a Greek foundation, where the Getic mixed with Greek elements is plausible, there is insufficient archaeological data to support this hypothesis.

Halmyris played an important strategic role during the Early Roman period (1st to 3rd centuries AD). Initially, an earth-fortification (last quarter of the 1st century AD), Halmyris was rebuilt in stone during the 2nd century AD as a fort by the military groups of the Legio I Italica and the Legio XI Claudia pia fidelis (Zahariade, 1986). Halmyris castrum had the most important role on the last segment of the Danubian limes, controlling the territory between Aegysus (Tulcea) and the mouth of Saint George or Dunavăţ arm (Suceveanu et al., 2003). Discovery of 2nd–3th century AD inscriptions on an altar dedicated to Hercules mention the existence of a settlement « vicus classicorum » (classicorum from Classis Flavia Moesia, the fleet organised by Emperor Vespasian). The dating of the epigraphic monuments (136 AD, 163 AD, 171 AD and 200 AD) suggests that during this period, a village founded by mariners from Classis Flavia Moesia was located near the fortress (Zahariade, 2012). This type of settlement is also known epigraphically from other provinces of the Roman Empire: vicus navaliorum (at Mainz, related to classis Germanica) and vicus portensium (at Nantes, related to civitas Namnetum) (Zahariade and Alexandrescu, 2011).

The Late Roman period of Halmyris spans an interval between the last quarter of the 3rd century AD and the 3rd or 4th decade of the 7th century AD (Suceveanu et al., 2003). At that time, the original rectangular shape of the fort was abandoned in favour of a triangular morphology (Zahariade and Phelps, 2002). The fort was composed of thick walls adjoined by numerous towers and three gates (Supplementary Material 1, available online; Mărgineanu-Cârstoiu and Apostol, 2015). The northern and the north-eastern gates were built towards the Danube. Civil constructions, such as thermae, suggest a change from primarily military to mixed civilian and military usage of the settlement in the 4th-century AD. Roman provincial pottery intermixed with Slavic coarse wares mark the presence of a non-Roman population in the fort during the second quarter of the 7th century AD and support the abandonment of the fortress by the Romans (Zahariade and Phelps, 2002).

Possible harbour location

Access to the city was possible from the sea by sailing back up the course of the Danube, which is believed to have flowed close to the Dunavăţ promontory in Late Antiquity (Zahariade and Phelps, 2002). Discovery of an inscription mentioning a « vicus classicorum » dated to the 2nd century AD suggests that during this period, close to the castrum, a civil settlement related to naval activities was founded by the discharged mariners of Classis Flavia Moesia (Zahariade and Alexandrescu, 2011). From the 4th-century AD, the harbour may have been a disembarkation point for military material, from large maritime vessels to fluvial ones as indicated by the ancient writer Zosimos (first half of the 5th century AD; Historia Nova, IV, 10). In the context of successive barbarian’s invasions during the 5th century AD archaeologists consider the fact that the military importance of Halmyris increased at that time (Suceveanu et al., 2003).

The presence of two natural decantation basins to the north of the site (Cruhlic Mic and Cruhlic Mare) are the relic of fluvial activity on this part of the floodplain (Figure 2). These palaeo-meanders are located at the southern limit of the inundation plain, at the foot of the Dunavăţ promontory in front of the major northern gate of the fort. According to Zahariade and Phelps (2002), this gate was mainly associated with harbour activities, and it was probably blocked due to the abandonment of the harbour during the second half of the 6th century AD.

Figure 2.

Location of Halmyris (ca. 2 km south of the Saint-Georges channel) and delimitation of the fluvial landforms observed north of the fortress. The paleogeography of the possible ancient channel surrounding the seasonal lake (Cruhlic Mare) is based on the observation of comparable landforms in other parts of the delta (e.g. Taranova lake). Image LandSat/Copernicus (slanting view).

Methods

Our work is based on the study of two sedimentary cores drilled on the Danube delta plain (HAI (335 cm in length): 45°1’32”N; 29°11’48”E; ca. + 1 m a.s.l.; and HAIII (577 cm in length): 45° 1’34”N; 29°11’56”E; ca. + 0 m a.s.l.) immediately near Halmyris (Figure 2 and Supplementary Material 1, available online). The cores, drilled using a percussion corer (Cobra TT), were attitudinally benchmarked relative to present mean sea level using a GPS. Core descriptions (texture, macrofauna content, organic remains) and sampling were undertaken during fieldwork.

Chironomid analyses of core HAI

Laboratory methods for the extraction and identiﬁcation of subfossil remains are described in Gandouin et al. (2005) and consisted of deﬂocculation with KOH, rinsing with water over a 100-µm sieve, and parafﬁn ﬂotation. We analysed 55 chironomid samples from core HAI (355 depth). A minimum of 50 head capsules per sample was chosen to provide statistically signiﬁcant estimates of environmental conditions (Heiri and Lotter, 2001). The identiﬁcation of head capsules was based on Brooks et al. (2007) and Klink and Moller Pillot (1999). Head capsules were identiﬁed under a stereomicroscope at 400X magniﬁcation. Chironomid diagrams were drawn using C2 version 1.7.2 (Juggins, 2007).

Statistical analyses have been performed with R Studio version 3.1.1: ade4, vegan and factoExtra were used for multivariate analyses (Chessel et al., 2004); rioja package for the constrained sum-of-squares cluster analysis (CONISS: in Juggins, 2015). Principal component analysis (PCA) was performed on the n (number of samples) by p (number of taxa) chironomid matrix of percentages. Data were square-root transformed to stabilise the variance. Rare taxa, that is, those present in only one sample or in less than 5% of all samples, were removed from the analysis.

Subfossil samples have been projected as passive objects for comparison between fossil and modern assemblages from Gandouin et al. (2006), in order to characterise the type of contemporary habitats. This allowed us to obtain an ordination of subfossil samples along a gradient of connectivity (materialised by scores on an axis of a between-class correspondence analysis) between the main channel and other floodplain habitats such as secondary channels, oxbow lakes, ponds and marshy environments.

Ostracods analyses and sedimentary texture of core HAIII

Bio-sedimentological analyses were performed following the methodology detailed in Marriner and Morhange (2007) and Marriner (2009) on 90 samples taken from core HAIII (577 cm depth). The general sediment texture, including the gravel (> 2 mm), sand (50 μm–2 mm) and silty-clay fractions (< 50 μm), was determined by wet sieving. Ostracods were picked from the fraction >160 μm and identified to species level using reference manuals and papers for Ponto-Caspian species (Boomer et al., 2010; Frenzel et al., 2010; Meisch, 2000; Opreanu, 2008; Tunoglu and Gökçen, 1997). To obtain reliable statistical estimates, we have picked around 300 valves per samples when the amount of sediments was sufficient.

Statistical analyses have been performed on the percentage matrix using the paleontological statistics software PAST (Version 2.14, Hammer et al., 2001). Rare species that represent less than 2% of the total individuals identified in this study were removed from the PCA. Ostracods were grouped according to three main assemblages; defined as shallow-marine, lagoonal and fresh/mesohaline inland waters. In order to test the ordination of samples by assessing major changes in palaeoenvironmental proxies, PCA was undertaken on the ostracods matrix.

Malacology of cores HAI and HAIII

Molluscs have been observed in both HAI and HAIII samples. Their identification was undertaken using Pfleger (1993). Due to the low number of individuals encountered in our study, the molluscan assemblage only permits to strengthen the information provided by the other biological proxies.

Dating of HAI and HAIII

The chronology is based on 15 AMS radiocarbon determinations performed at the Poznan Radiocarbon Laboratory (Poland) on plant remains, charcoal remains, bulk sediment, and articulated mollusc shells (Table 1). The dates obtained from articulated mollusc shells (Dreissena polymorpha and Cerastoderma sp.), have been corrected using a marine reservoir age. A marine reservoir age of 498 ± 41 (in Bony et al., 2015) has been subtracted from the radiocarbon age before calibration using the calibration curve IntCal13 (Reimer et al., 2013) in Clam (R Studio).

Table 1.

AMS-¹⁴C data expressed in calibrated years BP and BC at the 95% confidence level (2σ). b.s.: below surface, b.s.l.: below present mean sea level, a.s.l.: above present sea level. Calibration using R-code Clam (Blaauw, 2010) with the calibration curve IntCal13 (Reimer et al., 2013).

Sample	Laboratory number	Material	δC13	Depth (cm b.s.)	Depth (cm b.s.l.)	Age ¹⁴C	Age ¹⁴C (corrected)	2 sigma BP min; max	2 sigma BC/AD min; max	Status
HAI(55-60)	Poz-79628	Peat	−23.7	55–60	45–50 a.s.l.	110 ± 30 BP		12; 269	1681; 1938 AD	Accepted
HAI(157-163)	Poz-79629	Plan remains	−28.4	157–163	57–63	900 ± 30 BP		740; 911	1039; 1210 AD	Rejected
HAI(157-163)	Poz-79163	Mollusc shell (Cerastoderma sp.)	−2.5	157–163	57–63	2385 ± 30 BP	1785 ± 50	1569; 1823	127; 381 AD	Accepted
HAI(170-175)	Poz-79630	Peat	−24.3	170–175	70–75	2545 ± 30 BP		2498; 2749	800; 549 BC	Accepted
HAI(295-300)	Poz-79631	Peat	−26.6	295–300	195–200	4660 ± 30 BP		5315; 5467	3518; 3366 BC	Accepted
HAIII(60-63)	Poz-81693	Peat	−30.3	60–63	60–63	1230 ± 30 BP		1069; 1261	689; 802 AD	Rejected
HAIII(120-125)	Poz-79633	Plant remains	−26.3	120–125	120–125	830 ± 30 BP		688; 789	1161; 1262 AD	Accepted
HAIII(245-250)	Poz-81694	Organic sediment	−25.8	245–250	245–250	1930 ± 30 BP		1820; 1946	4; 130 AD	Rejected
HAIII(270-275)	Poz-79655	Charcoal	−29.1	270–275	270–275	1775 ± 30 BP		1611; 1812	138; 339 AD	Accepted
HAIII(275-280)	Poz-81695	Charcoal	−24.7	275–280	275–280	2585 ± 35 BP		1770; 2510	821; 561 BC	Accepted
HAIII(300-305)	Poz-79656	Charcoal	−32.9	300–305	300–305	3920 ± 35 BP		4242; 4496	2547; 2293 AD	Accepted
HAIII(355-360)	Poz-79657	Peat	−32.8	355–360	355–360	4425 ± 35 BP		4871; 5276	3327; 2922 BC	Accepted
HAIII(405-410)	Poz-81696	Peat	−30.2	405–410	405–410	5210 ± 40 BP		5903; 6174	4225; 3954 BC	Accepted
HAIII(470-480)	Poz-79659	Organic material	−30.6	470–480	470–480	5125 ± 35 BP		5749; 5939	3990; 3800 BC	Rejected
HAIII(530-540)	Poz-79164	Mollusc shell (Dreissena polymorpha)	−8	530–540	530–540	7170 ± 40 BP	6672 ± 57	7437; 7650	5701; 5488 BC	Accepted

With the retained radiocarbon determinations, we constructed an age–depth model for each core using the dedicated R-code Clam (Blaauw, 2010), which uses repeated random sampling of the calibrated distributions to derive a robust age–depth model. Thanks to these age–depth models, we obtained reliable ages for each unit and level analysed (figure presented in section 6.3).

Results

Faunal record and sedimentary texture of core HAIII

The ostracods identified are common in present-day ponds and lagoons of the Danube (Opreanu, 2003; Figure 3a). According to the ecology of the species, we differentiated three groups.

The first group includes endemic species of the Ponto-Caspian region (e.g. T. Amnicola, A. Bendovanica) associated with euryhaline marine environments (Boomer et al., 1996, 2010). Due to the stratification of Black Sea waters (Neretin et al., 2001), the surface waters are less saline (15–17‰) than the deep waters (23‰). Marine Mediterranean species are found at depths >20 m on the continental platform and Ponto-Caspian species are present in surface waters (Opreanu, 2005).

The ostracod Cyprideis torosa is an opportunistic species that can support holeuryhaline conditions (Boomer and Frenzel, 2011; Pint et al., 2012); it is the only species that constitutes the lagoonal assemblage.

The third assemblage comprises ostracods living in fresh to mesohaline inland water environments. Darwinula stevensoni (presented alone in Figure 3a) is associated with permanent and clear-waters. The other species are characteristic of stagnant waters (Candona neglecta, Cypria ophtalmica, Pseudocandona albicans).

The macrofauna assemblage is composed of brackish or lagoonal waters (Dreissena polymorpha) and freshwater species common to the Danube such as, Lithoglyphus naticoides, Theodoxus danubialis, Viviparus sp (Supplementary Material 2, available online).

Figure 3.

(a) Ostracod species and assemblages from the core HAIII. (b) Results of the PCA and reference samples. Plot of Axes 1 and 2 on the basis of ostracod associations.

By means of a PCA undertaken on the ostracod matrix, the different samples were reduced to two PCs that account for 78.8% of the total variability depending on their faunal contents. PC1 (64.3%) has high positive scores for samples composed of freshwater species and negative ones for those comprising marine species. PC2 (14.5%) allows us to categorise the freshwater species into two distinctive groups; samples with lotic species present negative scores and those containing freshwater species living in stagnant waterbodies present positive scores (Figure 3b).

The sedimentary sequence of core HAIII is typical of a regressive deltaic sequence with fluvial sediments overlying marine sediments. The sequence broadly displays three main textures; from fine-to-medium sands between ‒563 and ‒410 cm, silty-clay between ‒300 and ‒63 cm and then between ‒40 and the top of the core. Two organic-rich peat layers are intercalated in the sequence (between ‒410 and ‒300 cm and between ‒63 and ‒40 cm).

Faunal record of core HAI

Fifty-five chironomid samples were analysed. Twenty-six samples, mainly peat sediments, were devoid of head capsules (hc) or under the limit of 50 hc (Heiri and Lotter, 2001). Twenty-nine samples yielded 2080 identifiable hc (74 identified taxa). Among these samples, four stratigraphic levels were pooled (cf., 157–163, 180–190, 260–270 and 280–290 cm) in order to yield a significant number of hc. The cluster analysis allowed us to identify six chironomid zones: Hach-1 to Hach-6 (Figure 4a). Based on the typology of Gandouin et al. (2006), we differentiated three ecological groups: lentic, ubiquitous and lotic taxa. The main chironomid taxa and their dynamics are summarised in Table 2.

Figure 4.

(a) Chironomid species and assemblages. (b) Results of the PCA and reference samples. Plot of Axes 1 and 2 on the basis of chironomid associations.

Table 2.

Main chironomid taxa and their dynamics over the HAI record.

Faunal zones	Main assemblage dynamics	Lithology
Hach-6(50–35 cm)	Increasing percentages of Glyptotendipes and Dicrotendipes nervosus.	Peaty sediments
Hach-5(95–50 cm)	Dominance of Polypedilum.Increase in Procladius percentages.	Clayey sediment
Hach-4(145–95 cm)	Dominance of Cricotopus and Dicrotendipes notatus.First appearance of Eukiefferiella/Tvetenia, Rheotanytarsus, Micropsectra and Neozavrelia.	Sandy silt sediment
Hach-3(163–145 cm)	First appearance of Psectrocladius, Halocladius, Microchironomus and Acricotopus.	Sandy silt sediment
Hach-2(300–163 cm)	Dominance of Chironomus, Dicrotendipes notatus, Glyptotendipes. Significant abundances of Endochironomus and Einfeldia.	Peaty sediments
Hach-1(335–300 cm)	Dominance of Dicrotendipes nervosus and Polypedilum.Highest abundance over the record of Procladius, Cryptochironomus and Harnischia.	Sandy silt sediments

The PCA axis 1 explains 17.9% of the total variance of the data set, 5% for PCA axis 2. The first axis shows a clear contrast between two groups of taxa. The first group (positive scores) is mainly characterised by Chironomus, Dicrotendipes, Polydedilum and Phaenopsectra. The second group (negative scores) is characterised by Psectrocladius, Cladotanytarsus and Cricotopus. Procladius strongly contributes to the positive side of the PCA axis 2. The positive PCA scores comprise mostly peaty sediments (particularly samples 273 and 285 cm), while the negative scores constitute sandy sediments (155 cm in particular). Clayey (73, 78 and 83 cm) and silty sediments (323, 328 and 333 cm) comprise positive PCA axis 2 scores. PCA analysis demonstrates a strong opposition between lotic or ubiquitous assemblages such as (e.g., Cricotopus, Cladotanytarsus, Neozavrelia) and lentic ones (e.g. Dicrotendipes notatus, Glyptotendipes, Limnophyes and Phaenopsectra), which is probably induced by the transversal connectivity gradient between the main fluvial channel and other floodplain waterbodies.

The passive projection of subfossil samples into the ordination of the between-class Correspondence Analysis (Supplementary Material 3, available online) performed by Gandouin et al. (2006) shows that the fossil data set are close to lotic stations such as connected side arms (SA1 et SA2) and the main channel (MCRh). Two groups of samples can be identified. The more organic samples are close to the station SA2, a temporarily connected side arm (Garcia and Laville, 2001), while minerogenic samples (sand, silt and clay) are close to a permanently connected side arm (SA1) and the MCRh.

Two hundred and thirty-five mollusc shells have been identified in core HAI (Supplementary Material 2B, available online). In silty dominated samples, between 280 and 320 cm, a large number of freshwater and stagnant gastropods (such as Planorbis planorbis and Planorbis corneus) were found. Sandy silts from 145 to 165 cm were characterised by an abundance of halotolerant (Dreissena polymorpha) and lagoonal (Cerastoderma glaucum) bivalves. In sandy peat from 95 to 145 cm deep, a majority of Dreissena polymorpha was found. Finally, the more clayey samples (between 70 and 95 cm) were marked by the presence of Lithoglyphus naticoides, a taxon inhabiting slow-flowing and muddy environments.

Cores chronology

Fifteen radiocarbon-dated samples were taken from various depths, providing a chronological framework from 7170 to 110 uncalibrated years BP (Table 1). Among them, we identified some age reversals. In core HAI, we removed the date, Poz-79629 (900 ± 30 BP; 1039–1210 cal. yr AD) from the age–depth model because of the presence of roots in the plant remains observed during the sampling preparation. In core HAIII, at the bottom of the core (470–480 cm depth), the date Poz-79659 (5125 ± 35 BP; 3990–3800 cal. yr BC) was rejected due to possible reworking. This date was obtained from organic material collected in the unit identified as the channel/river mouth deposits and was possibly eroded from upstream and transported by the river. Above the first peat layer, we chose to reject the date Poz-81694 (1930 ± 30 BP; 4–130 cal. yr AD; 245–250 cm depth) because we dated the bulk sediment. We have chosen to use the dates performed on organic remains taken from the same unit. At the top of the core (60–63), the bottom of the second peat layer (Poz-81693) seems to be too old (1230 ± 30 BP; 689–802 cal. yr AD) compared to the dating of the peat layer of core HAI, dated to 110 ± 30 BP (1681–1938 cal. yr AD; Poz-79628; 55–60 cm depth).

Mean sedimentation rates calculated for the cores HAI and HAIII are summarised in Table 3. The age–depth models (Supplementary Material 4, available online) reveal a possible sedimentary hiatus at the top of the peat layer (between units Hach-2 and Hach-3 in HAI and units D and E in HAIII).

Table 3.

Mean sedimentation rates calculated for cores HAI and HAIII.

Core	Cores section in cm from core top	Calculated mean sedimentation rates
HAI	57.5–160	0.61 mm.yr⁻¹
HAI	160–172.5	0.12 mm.yr⁻¹
HAI	172.5–297.5	0.43 mm.yr⁻¹
HAIII	122.5–272.5	1.34 mm.yr⁻¹
HAIII	272.5–277.5	0.18 mm.yr⁻¹
HAIII	277.5–302.5	0.09 mm.yr⁻¹
HAIII	302.5–357.5	0.63 mm.yr⁻¹
HAIII	357.5–407.5	0.43 mm.yr⁻¹
HAIII	407.5–535	0.78 mm.yr⁻¹

Discussion

In the previous section, the results of the two cores were presented using a metric scale. Here, in order to compare the results of the two cores, we have chosen to represent the main bio-sedimentological units on a chronological scale in order to place the two cores in a broader palaeoenvironmental context (Figure 5). Because core HAIII represents a more complete sedimentary sequence than HAI, we decided to discuss the palaeoenvironmental evolution of Halmyris based on the main bio-sedimentological units identified in HAIII. The chironomid data from core HAI are used to reinforce our interpretations.

Figure 5.

Comparison of the biosedimentological units of cores HAI and HAIII on a chronological timescale.

Palaeoenvironmental phases recorded in cores HAI and HAIII

Unit A: Marine bay before ca. 7550 cal. yr BP (ca. 5600 cal. yr BC)

The top of this unit, in HAIII, is dated to ca. 7550 cal. yr BP (ca. 5600 cal. yr BC). It is composed of medium to fine yellow/orange sands. Macrofauna is represented by the species Dreissena polymorpha and Theodoxus danubialis, which are fluvial species endemic to the Pontic region and very frequently encountered in the Danube riverbed. The samples are dominated by lagoonal and euryhaline marine ostracods and reflect negatives values of the PCA Axis-1 (Figure 3). The combination of marine ostracods endemic to the Black Sea (e.g. Tyrenocythere amnicola) with the opportunistic species Cyprideis torosa mark the presence of the sea after the reconnection of the Black Sea to the global ocean ca. 9400 cal. yr BP (Soulet et al., 2011). This unit presents a marine (coastal) sedimentation in a context of high freshwater supply, in front of the mouth of the Danube.

Unit B: Delta-front deposits between ca. 7550 cal. yr BP (ca. 5600 cal. yr BC) and ca. 6680 cal. yr BP (ca. 4730 cal. yr BC)

The rapid progradation of the Old Danube lobe led to the deposition of medium grey sands. Only two freshwater species were identified for the macrofauna (Dreissena polymorpha and Viviparus sp.). The absence of ostracods in the sand could be linked to high sedimentation rates and/or high fluvial energy, evoking the position of the delta front as proposed by Vespremeanu-Stroe et al. (2017) (Figure 1b).

Unit C: Lower delta plain environment between ca. 6680 cal. yr BP (ca. 4730 cal. yr BC) and ca. 6040 cal. yr BP (4225–3954 cal. yr BC)

We observe a decrease in the grain-size probably linked to a decrease in river flow energy. Ostracod fauna is present in two samples and comprises a mixture of fresh to euryhaline species commonly found in deltaic environments (e.g. Candona neglecta, Heterocypris salina). The end of this unit corresponds to the first chironomid zone Hach-1, from ca. 6200 cal. yr BP to ca. 5390 cal. yr BP. The abundance of Dicrotendipes nervosus suggests a developed macrophytic vegetation (Brodersen et al., 2001). Abundances of Polypedilum indicate eutrophic waters (Klink, 2002). The presence of Harnishia, which is nowadays associated with large waterbodies with stagnant or slow-flowing water on a mineral bottom (Moller Pillot and Buskens, 1990), coupled with the presence of Cryptochironomus, indicate a sandy substrate mixed with fine organic matter (Vallenduuk and Morozova, 2005). These data are in agreement with the stratigraphy. Some species of Cryptochironomus are strictly predatory (Armitage et al., 1995), which is also the case for Procladius. This latter is also an oxy-regulator taxa, and it is tolerant to daily water-oxygen fluctuations, due to intense photosynthetic activities (Brodersen et al., 2004). The passive projection of Halmyris fossil data on the modern dataset ordination (Gandouin et al., 2006) suggests a permanently connected side arm. Freshwater molluscs (e.g. Anisus vortex, Planorbis planorbis and Lymnae) from core HAI are in agreement with chironomids and ostracods (Supplementary Material 3, available online) and the prevalence of slow flowing freshwater conditions. The decrease in the grain size in comparison to the previous unit could be explained by the displacement of the channel from the palaeo-cliff, possibly translating a fluvial avulsion. Channel avulsions are common on fluvial-dominated deltas and play an important role in their morphogenesis (Jones and Schumm, 1999; Stouthamer et al., 2011).

Unit D: Peat layer between 5210 ± 40 BP (4225–3954 cal. yr BC) and 3920 ± 35 BP (2547–2293 cal. yr BC)

In HAIII, this organic-rich peat layer formed between 5210 ± 40 BP (4225–3954 cal. yr BC) and 3920 ± 35 BP (2547–2293 cal. yr BC). Vespremeanu-Stroe et al. (2017) have demonstrated that the formation of peat is common in this area (between the Saint George arm and the Razelm-Sinoe lagoon). Formation of such peat layers reflects the very low sedimentary inputs into the inner delta at this time, due to the important progradation of the Saint-George lobe and the export of the sediment to the eastern delta and the Black Sea shoreline.

This unit corresponds to the second chironomid zone Hach-2 from 4660 ± 30 BP (3620–3363 cal. yr BC) and 2545 ± 30 BP (800–550 cal. yr BC), with the sporadic presence of chironomid and mollusc subfossils that reveal a probable disconnection of the site from the fluvial system, possibly leading to conservation problems. Projection of these results on the modern data set (Gandouin et al., 2006) evokes a temporarily connected side channel with vegetated and eutrophic waters as suggested by Chironomus, Dicrotendipes notatus, D. nervosus and Glyptotendipes (probably G. pallens). Dicrotendipes notatus is presently associated with a very slow flowing channel from the Danube River with numerous dead-leaves on the river bottom (Moller Pillot and Buskens, 1990). Nowadays, G. pallens larva thrive in submerged woods and plants or build transportable cases on solid substrates. Occasionally, they are found in slightly brackish waters (Vallenduuk, 1999). Throughout the zone, both PCA axis 1 and between-class CA scores show an increasing trend towards higher connectivity of the site with the main channel. Particularly from ca. 2800 cal. yr BP (180 cm), which is contemporaneous with the foundation of Halmyris (about 650 cal. yr BC).

Unit E: Distributary channel in Getic and Roman times (ca. 6th century BC to ca. 7th century AD)

In the two cores, this unit is located above the peat layer, and it is composed of silty-sand that incorporate significant freshwater shell debris. In HAIII, this unit is dated after 3920 ± 35 BP (2547–2293 cal. yr BC; date obtained on the top of the peat layer). The three ages obtained in this unit highlight an important chronological gap (Table 1 and Supplementary Material 4, available online). This chronological gap may be due to the erosion of the peat layer as a result of fluvial activity (or anthropogenic dredging).

The macrofauna is composed of Dreissena polymorpha living in rivers and Lithoglyphus naticoides and Viviparus sp., living on the riverbanks of low-energy rivers (Pfleger, 1993). The ostracods are composed of a mixture of species that have a wide ecological range including lagoons, springs, ponds and lakes (Frenzel et al., 2010; Opreanu, 2003; Salel et al., 2016). The presence of Darwinula stevensoni, which does not tolerate drying up of the water body, testifies to permanent submerged conditions. Furthermore, this species is a lotic ostracod associated with fluvial interstitial sands (Dole-Olivier et al., 2000). At the base of the unit, samples are dominated by Darwinula stevensoni and reflect negatives values of the PCA Axis-2, consistent with ostracods living in permanent waters (Figure 3b). This facies could be identified as a secondary channel dating from the Getic/Greek (2585 ± 35 BP; 820–566 cal. yr BC; 275–280 cm depth) and the Roman periods (1775 ± 30 BP; 138–339 cal. yr AD; 270–275 cm depth), with moderate flowing waters, allowing the development of a large population of ostracods (ca. 500 valves for 20 g of sediment). This channel may have formed after an avulsion of the main course of the Danube, which can occur rapidly in deltaic environments (Jones and Schumm, 1999). Progressively, the connection with the main fluvial channel decreases and the samples are dominated by the ostracod Candona neglecta that is characteristic of stagnant water bodies (positive PCA Axis-2 scores; Fuhrmann, 2012). Furthermore, the proportion of Darwinula stevensoni is still >5% up to 220 cm depth and shows that the secondary channel remains connected to the main channel (Figure 3a).

Chironomids confirm this hypothesis, with high scores of both PCA axis 1 and between-class CA, characteristic of a permanently connected side-arm (during Hach-3 and 4). In the sandy zone Hach-3 (163–145 cm depth) dated between ca. 210 cal. yr BC and ca. 202 cal. yr AD, we found high percentages of Chironomus and Glyptotendipes suggesting the persistence of eutrophic waters. These taxa, in association with halotolerant species, such as Halocladius and Psectrocladius sordidellus (Klink and Moller Pillot, 1999), as well as the presence of numerous shells of halotolerant (Dreissena polymorpha) and lagoonal (Cerastoderma glaucum) bivalves, point to increasing salinity at the site.

In zone Hach-4 (145–95 cm), the dominance of Cricotopus and the appearance of several lotic taxa such as Eukiefferiella/Tvetenia, Rheotanytarsus and Micropsectra, suggest that the site was always connected to the main channel. Neozavrelia has always been found in association with Nuphar luteola (Thienemann, 1942) and other bryophytes in artificial riffles from the lower part of the Rhine river (Klink, 2002).

Unit F: Residual channel during the 6th–7th centuries AD

In HAIII, the unit is dated between ca. 1360 cal. yr BP (ca. 590 cal. yr AD) and 1230 cal. yr BP (ca. 720 cal. yr AD). It is composed of fine sediments (silts and clay > 98 %), as is generally the case for residual channels (Toonen et al., 2012). The change in the dominant ostracod species (from Darwinula stevensoni to Candona sp.) underscores the shift from a connected-fluvial channel to a probable stagnant water body. In HAI, scores show an abrupt decrease around 140–130 cm, corresponding to 1400–1245 cal. yr BP (550–700 cal. yr AD), highlighting a probable temporary disconnection towards the end of the Roman occupation of Halmyris, which is contemporaneous with the disconnection recorded in core HAIII (Figure 5).

Unit G: Upper delta plain between 1230 cal. yr BP (ca. 720 cal. yr AD) and 110 ± 30 BP (1681–1938 cal. yr AD)

This unit, located between 63 and 200 cm depth, started to accrete after 1230 cal. yr BP (ca. 720 cal. yr AD) in core HAIII and is dated at its centre to 830 ± 30 BP (1160–1260 cal. yr AD). In core HAI, this unit corresponds to the chironomid zone Hach-5 dated between ca. 1300 cal. yr AD and 110 ± 30 BP (1681–1938 cal. yr AD). It is mostly composed of silts and clay (85%–98%) and the sedimentation is very homogeneous. The absence of aquatic fauna on HAIII could demonstrate the transformation of the area into a dried-out floodplain because of the migration of the river. The sediment deposition may translate a succession of overbank flooding. The chironomid data in HAI show a trend towards a fluvial-disconnection as suggested by a progressive decrease in PCA axis-1 values and between-class CA scores over Hach-5, which is due to the increase of lentic taxa such as Chironomus, Dicrotendipes and Glyptotendipes (Gandouin et al., 2006). The presence of Lithoglyphus natocoides, indicative of stagnant or slow-flowing freshwater, meshes with this hypothesis.

Unit H: Organic-rich peat layer

The development of a second organic-rich peat layer attests to the presence of freshwater inputs that favoured peat growth during the 19th century. These inputs of freshwater could be linked to a rise in the water table in this area (groundwater flows). The chironomid content in the peat layer of core HAI demonstrates that the level of connection to the main river was poor, with the disappearance of lotic taxa. The development of this peat layer, during the Little Ice Age, seems to underscore a reduction in sediment inputs near Halmyris in the period of generally high fluvial activity due to secular climatic degradation (McCarney-Castle et al., 2012). This reduction could be because the main flow of the delta was concentrated in the Chilia lobe (northern part of the delta; Filip and Giosan, 2014; Vespremeanu-Stroe et al., 2017). This disconnection could also be due to the containment of the Danube River, which became widespread in the mid-nineteenth century in the Danube area (Gupta, 2007) and elsewhere in Europe (Tockner et al., 2009).

Geoarchaeological implications and characterisation of the anchorage

The comparison of cores HAI and HAIII reveals the presence of a fluvial channel north of the settlement of Halmyris, attested by the presence of lotic ostracods and chironomid species (Figures 3 –5). We have demonstrated that it was active during the occupation of Halmyris (6th century BC–7th century AD). The presence of the channel from the middle of the second millennium BC could in part explain the choice of this emplacement for the location of the Getic settlement and the possible Greek emporium in place for the later fortress in the 6th century BC.

Navigation in the channel – Water depth

To ascertain if this channel was navigable, we need to know its nautical dimensions. In particular, the depth of the channel will help us to elucidate the maximum draught of the ships (Boetto, 2010; Salomon et al., 2016). For this purpose, we based the reconstruction on the core HAIII because the core HAI was drilled close to the palaeo-bank of the channel and, as a result, the water depth in this area was much lower. Because the ancient harbour of Halmyris was situated on the deltaic plain of the Danube, ca. 38 km from the present shoreline (but closer to the river mouth during Antiquity), the water level of the channel is linked to the elevation of the sea level. It is largely assumed that the post-glacial sea-level rise in the Black Sea is comparable to that of the Mediterranean, comprising a rapid glacio-eustatic rise until 6000 BP, followed by an important deceleration in sea-level rise that reached its present position around 2000 years ago (Brückner et al., 2010; Giosan et al., 2006). Nevertheless, we have no data precisely constraining the level of the Black Sea around 2000 years ago for the Danube delta. Using the ICE-5G (VM2) GIA model (Peltier, 2004), we have constrained the RSL position of the Black Sea for the two periods of human occupation at Halmyris. We propose the related water depth of the channel (Table 4 and Figure 6). In the 6th century BC (Getic/Greek occupation of Halmyris), the level of the Black Sea was between 76 and 44 cm below the present mean sea level and the water column of the channel was at least 163 ± 9 cm deep (mean 244 ± 9 cm). At the end of the occupation of Halmyris, the level of the Black Sea was between 29 and 22 cm below the present mean sea level and the water column of the channel was at least 155 ± 9 cm deep (mean 236 ± 9 cm).

Table 4.

Estimate of the water depth of the channel at different times. The modern tidal range is from Medvedev et al. (2016). Maximum and minimum water depth at Sulina mouth from Bondar and Iordache (2016). Mean water depth = (A + (B + C)); Minimum water depth = (A + (B + D)); where A = Depth of the sample; B = Minimum sea-level position; C = Average level Sulina mouth (1840–2011); D = Minimum level Sulina mouth 1921.

Dating of the channel (14 C)	Datingof the channel (cal. BC/AD)	Depth of the sample (cm b.s.l.)	Minimal sea-level position (modelled)	Mean sea-level position (modelled)	Modern tidal range (cm)	Maximum level Sulina mouth (May 2006)	Average level Sulina mouth (1840-2011)	Minimum level Sulina mouth (January 1921)	Mean water depth of the channel (cm)	Minimum water depth of the channel (cm)
(age depth model)	600–650 AD	220	−29 cm	−22 cm	18	+ 137 cm	+ 45 cm	−36 cm	236 ± 9	155 ± 9
1775 ± 30	140–340 AD	270–275	−44 cm	−27 cm	18	+ 137 cm	+ 45 cm	−36 cm	271 ± 9	190 ± 9
2585 ± 35	820–565 BC	275–280	−76 cm	−44 cm	18	+ 137 cm	+ 45 cm	−36 cm	244 ± 9	163 ± 9

Figure 6.

Reconstruction of the channel depth and sea-level position for the Roman period (600 AD) in core HAIII associated with the main stratigraphic units, the sedimentation rate and an example of the boat used in Late Roman times on the Rhine river.

The channel was more than deep enough for the circulation of flat-bottomed vessels typically used on rivers during Antiquity. In Roman times, the harbour of Halmyris was mainly used for military purposes. Bounegru and Zahariade (1996) present different types of fluvial boats used between the 1st and the 6th centuries AD on the Lower Danube. Although they provide little information regarding the draught of boats used between the 1st and the 3rd centuries AD, they do state that they were flat-bottomed boats adapted to fluvial navigation such as liburnae (a liburna has been identified in a text from the 2nd century AD in Noviodunum, another fortress of the Danubian limes, situated ca. 60 km upriver from Halmyris; Bounegru and Zahariade, 1996). The fleet of the Lower Danube was particularly important at the beginning of the 5th century AD and comprised, according to the Theodosian code (7.17.1. January 412), 225 lusoriae (Syvanne, 2015). Navis lusoriae were war-ships intended for fluvial navigation on the Frontier Rivers of the Roman Empire (Pitassi, 2011; Torr, 1894). The use of these ships is attested by a 4th-century Roman author, Vegetius (Epitoma Rei Militaris, 4.34, 4.46) that mentions small vessels, including navis lusoriae, that were used on the Danube River (Syvanne, 2015). Such boats, intended to protect the borders of the Empire, were also used on the Rhine River. Archaeological excavations undertaken in Mainz (Mogontiacum) have provided important information about river vessels used in the Late Roman period. In the ancient harbour basin of Mogontiacum, archaeologists have discovered several shipwrecks dated to the 3rd and the 4th centuries AD (Höckmann, 1993). Two different types of boats have been identified, navis lusoriae (Mainz 1 and 5) and a smaller vessel used as a patrol ship for the surveillance of the Rhine border during the 4th century (NAVIS Project, Römisch-Germanisches Zentralmuseum, Mainz). According to reconstructions of the boats by archaeologists at the ‘Museum of Ancient Shipping’ in Mainz (Supplementary Material 5, available online), it is noted that such fluvial military vessels had a small draught (around 1 m) and may have been used at Halmyris.

Relationship between settlements and rivers in ancient times

Our study shows that Halmyris’ inhabitants exploited a natural fluvial channel to host their harbour in Getic/Greek and Roman times. In Antiquity, settlements located along river channels used riverine flow which explains why several of them have densely artificialized riverbanks, particularly in urban areas (Allinne, 2007). Numerous rural settlements were founded along the palaeo-channels of the Rhone delta (France), particularly between the 3rd century BC and the 7th century AD. Arnaud-Fassetta and Landuré (2014) have demonstrated that these settlements were mostly located along the main channels of the river and that human occupation was coeval with the period of activity of the channels. This link is also visible on the Nile delta, where the ancient city of Pelusium, originally located at the mouth of the Pelusiac branch of the Nile, seems to have been abandoned following the shift of the river channel to the east and the infilling of the initial river-mouth due to the progradation of the coastline (Goodfriend and Stanley, 1999; Stanley et al., 2008). The same phenomenon occurred in the Rhine delta, where van Dinter et al. (2017) have highlighted a strong relationship between settlement history and river activity, furthering addition to other socio-economic and political factors. Halmyris was an important Roman fortress installed along the limes of the Lower Danube. The identification of a natural harbour on the banks of a secondary channel leads us to draw comparisons with other fluvial harbours present in the region. Upstream, on the southern bank of the Danube, the fortress of Capidava has furnished interesting information regarding harbour infrastructure (Munteanu, 2012). Archaeological investigations have unearthed an artificial quay 2.5 m wide and 64 m long to the northwest of the fortress (Dobrinescu and Bodolică, 2016; Supplementary Material 6, available online). This harbour was used between the 2nd and the 4th centuries AD. The structure, in a perfectly upright position, is fixed on the rocky substratum. At Halmyris, no archaeological surveys in the supposed harbour area have been undertaken. The ancient harbour could have been managed as the harbour of Capidava because the channel, in front of the main gate, is conducive harbour activity.

Conclusion

The Roman limes of the lower Danube was protected by a series of fortresses and harbours. Our study demonstrates the presence of a channel to the north of the fortress of Halmyris which confirms the previous archaeological hypotheses regarding harbour activities in this area. Furthermore, it provides useful information to understand the interplay between the natural environment and the organisation of ancient harbours in deltaic contexts. The harbour of Halmyris was probably located in the reconstructed channel that flowed in close proximity to the site until the 7th century AD. The fortress enjoyed an easy access to the river while being protected from the floods of the Danube due to its position on the promontory (palaeo-cliff). We have demonstrated that this channel was navigable throughout the period when the site was occupied (Getic/Greek and Roman Periods). At the time of the abandonment of Halmyris in the 7th century AD, the depth of the channel was > 155 cm, allowing the circulation of fluvial-military boats. Zahariade and Phelps (2002) have arbitrarily attributed the complete obstruction of the northern gateway to the fortress during the second half of the 6th century to the navigation difficulties encountered to reach Halmyris. Our core reveals that the disconnection of the channel of Halmyris from the main channel was contemporaneous with the abandonment of the site. However, we cannot determine whether the abandonment of the site resulted in a natural disconnection between the secondary channel of Halmyris and the main channel. A fall in population, associated with the general geopolitical situation of the Late Roman Empire, could also have resulted in poor maintenance of the connection between the main channel and the secondary channel.

Supplemental Material

Supplementary_data_Halmyris – Supplemental material for Halmyris: Geoarchaeology of a fluvial harbour on the Danube Delta (Dobrogea, Romania)

Supplemental material, Supplementary_data_Halmyris for Halmyris: Geoarchaeology of a fluvial harbour on the Danube Delta (Dobrogea, Romania) by Matthieu Giaime, Gwenaël Magne, Alexandra Bivolaru, Emmanuel Gandouin, Nick Marriner and Christophe Morhange in The Holocene

Footnotes

Acknowledgements

M. Giaime acknowledges the support of the Institute of Advanced Studies and the Department of Geography at Durham University and was supported by a Durham Junior Research Fellowship, cofunded by the European Union under grant agreement number 609412. We thank M. Vacchi (University of Exeter) who provided the sea-level model for the Black Sea. We express our gratitude to V. Lungu, M. Streinu and the Museum of Tulcea for the work permits, as well as to M. Zahariade, director of the archaeological excavations at Halmyris. The authors would like to thank Dr. H. Bruneton (Aix-Marseille Université), Dr. P. Carbonel (CNRS), Dr. V. Ollivier (CNRS-LAMPEA) and two anonymous reviewers for their constructive remarks on an earlier version of the manuscript. M. Giaime acknowledges MOPP-Medflood (INQUA CMP project 1603 P) for fruitful discussions.

Funding

The project leading to this publication received funding from the Excellence Initiative of Aix-Marseille Universite–A*MIDEX, a French ‘Investissements d’Avenir’ project in the framework of Labex OT-Med (ANR-11-LABEX-0061). Support was also provided by the Institut Universitaire de France and the Fondation Eccorev (Aix-Marseille Universite).

ORCID iD

Matthieu Giaime

References

Allinne

(2007) Les villes romaines face aux inondations. La place des données archéologiques dans l’étude des risques fluviaux. Géomorphologie: Relief, Processus, Environnement 13(1): 67–84.

Allinne

Morhange

Pasquinucci

et al . (2016) Géoarchéologie des ports de Pise « Stazione Ferroviaria San Rossore » et de Portus Pisanus, dynamiques géomorphologiques, sources antiques et données archéologiques. Revue Archéologique de Narbonnaise 44: 321–338.

Anthony

Marriner

Morhange

(2014) Human influence and the changing geomorphology of Mediterranean deltas and coasts over the last 6000 years: From progradation to destruction phase? Earth-Science Reviews 139: 336–361.

Antipa

(1914) Citeva probleme stiintifice si economice privitoare la Delta Dunarii. Analele Academiei Române. Memoriile Secţiunii Ştiinţifice II 36: 61–135.

Armitage

Cranston

Pinder

(1995) The Chironomidae: The Biology and Ecology of Non-Biting Midges. London: Chapman & Hall.

Arnaud-Fassetta

Landuré

(2014) Le risque fluvial en milieu rural de l’époque grecque au haut Moyen Âge, Le cas du delta du Rhône (France). In: Carcaud

Arnaud-Fasseta

(eds) La géoarchéologie française au XXIème Siècle. Paris: CNRS, pp. 215–236.

Avram

Hind

Tsetskhladze

(2004) The Black Sea area. In: Hansen

Nielsen

(eds) An Inventory of Archaic and Classical Poleis. Oxford: Oxford University Press, pp. 924–973.

Benvenuti

Mariotti-Lippi

Pallecchi

et al . (2006) Late-Holocene catastrophic floods in the terminal Arno River (Pisa, Central Italy) from the story of a Roman riverine harbour. The Holocene 16(6): 863–876.

Bivolaru

Giaime

Bottez

et al . (in press) Istros (Black Sea coast, Romania) - A geoarchaeological perspective regarding harbour(s) location. In: Blue

Sivan

Farr

et al . (eds) Proceeding of The Honor Frost Foundation Conference on Mediterranean Maritime Archaeology. Leiden: Sidestone Press.

10.

Blaauw

(2010) Methods and code for ‘classical’ age-modelling of radiocarbon sequences. Quaternary Geochronology 5: 512–518.

11.

Boetto

(2010) Le port vu de la mer: l’apport de l’archéologie navale à l’étude des ports antiques. In: Keay

Boetto

(eds) Portus, Ostia and the Ports of the Roman Mediterranean. Contributions from Archaeology and History. Rome: Ministero per i Beni e le Attività Culturali, pp. 112–128.

12.

Bondar

Iordache

(2016) Sediment transport on the Romanian section of the Danube River. Geo-Eco-Marina 22: 29–49.

13.

Bony

Baralis

Lungu

et al . (2013) Mobilité des paysages et stratégies coloniales au sud du delta du Danube: la colonie grecque d’Orgame/Argamum (Jurilovca, département de Tulcea, Roumanie) In: Saint-Martin

(ed.) Recherches croisées en dobrogea. Bucharest: Editura Amanda Edit, pp. 133–156.

14.

Bony

Morhange

Marriner

et al . (2015) History and influence of the Danube delta lobes on the evolution of the ancient harbour of Orgame (Dobrogea, Romania). Journal of Archaeological Science 61: 186–203.

15.

Boomer

Frenzel

(2011) Possible environmental & biological controls on carapace size in Cyprideis torosa (Jones, 1850). Joannea: Geologie und Palaontologie 11: 26–27.

16.

Boomer

Guichard

Lericolais

(2010) Late Pleistocene to Recent ostracod assemblages from the western Black Sea. Journal of Micropalaeontology 29(2): 119–133.

17.

Boomer

Whatley

Aladin

(1996) Aral Sea Ostracoda as environmental indicators. Lethaia 29(1): 77–85.

18.

Bounegru

Zahariade

(1996) Les Forces Navales du Bas Danube et de la Mer Noire aux Ier–VIe siècles. In: Tsetkhladze

(ed.) Colloquia Pontica 2. Oxford: Oxbow Books, 124 pp.

19.

Brodersen

Pedersen

Lindegaard

et al . (2004) Chironomids (Diptera) and oxy-regulatory capacity: An experimental approach to paleolimnological interpretation. Limnology and Oceanography 49: 1549–1559.

20.

Brooks

Langdon

Heiri

(2007) The Identification and Use of Palaearctic Chironomidae Larvae in Paleoecology. London: Quaternary Research Association.

21.

Brückner

Herda

Müllenhoff

et al . (2014) On the Lion Harbour and other harbours in Miletos: Recent historical, archaeological, sedimentological, and geophysical research. Proceedings of the Danish Institute at Athens 7(7): 49–103.

22.

Brückner

Kelterbaum

Marunchak

et al . (2010) The Holocene sea level story since 7500 BP – Lessons from the Eastern Mediterranean, the Black and the Azov Sea. Quaternary International 225(2): 160–179.

23.

Brückner

Müllenhoff

Gehrels

et al . (2006) From archipelago to floodplain–geographical and ecological changes in Miletus and its environs during the past six millennia (Western Anatolia, Turkey). Zeitschrift für Geomorphologie 142: 63–83.

24.

Burchfiel

Bleahu

Borcos

et al . (1974) Geology of Romania. Geology 2(8): 392–394.

25.

Carozza

Micu

Mihail

et al . (2012) Landscape change and archaeological settlements in the lower Danube valley and delta from early Neolithic to Chalcolithic time: A review. Quaternary International 261: 21–31.

26.

Carozza

Micu

Carozza

et al . (2013) Le tell submergé chalcolithique de Taraschina et l’évolution interne du delta du Danube – Regards croisés à partir des données archéologiques et géo-archéologiques. In: Saint-Martin

(ed.). Recherches croisées en Dobrodgea. Bucharest: Editura Amanda Edit, pp. 97–119.

27.

Chessel

Dufour

Thioulouse

(2004) The ade4 package-I-One-table methods. R News 4: 5–10.

28.

Dobrinescu

Bodolică

(2016) Capidava 2015 – Sector PORT, Debarcader/ Molul portuar antic (Cronica Cercetărilor Arheologice Din România). Report: Campania 2015. București: Institutul Național Al Patrimoniu, pp. 155–157. Available at: http://cronica.cimec.ro/detail.asp?k=5587&d=Capidava-Topalu-Constanta-Port-2015.

29.

Dole-Olivier

Galassi

DMP

Marmonier

et al . (2000) The biology and ecology of lotic microcrustaceans. Freshwater Biology 44(1): 63–91.

30.

Filip

Giosan

(2014) Evolution of Chilia lobes of the Danube Delta: Reorganization of deltaic processes under cultural pressure. Anthropocene 5: 65–70.

31.

Fitzsimmons

Hambach

(2014) Loess accumulation during the last glacial maximum: Evidence from Urluia, southeastern Romania. Quaternary International 334: 74–85.

32.

Flaux

El-Assal

Shaalan

et al . (2017) Geoarchaeology of Portus Mareoticus: Ancient Alexandria’s lake harbour (Nile Delta, Egypt). Journal of Archaeological Science: Reports 13: 669–681.

33.

Frenzel

Keyser

Viehberg

(2010) An illustrated key and (palaeo) ecological primer for Postglacial to Recent Ostracoda (Crustacea) of the Baltic Sea. Boreas 39(3): 567–575.

34.

Fuhrmann

(2012) Atlas quartärer und rezenter Ostrakoden Mitteldeutschlands. Altenburg: Naturkundliches Museum Mauritianum.

35.

Gandouin

Franquet

Vliet-Lanoë

(2005) Chironomids (Diptera) in river floodplains: Their status and potential use for palaeoenvironmental reconstruction purposes. Archiv für Hydrobiologie 162(4): 511–534.

36.

Gandouin

Maasri

Van Vliet-Lanoe

et al . (2006) Chironomid (Insecta: Diptera) assemblages from a gradient of lotic and lentic waterbodies in river floodplains of France: A methodological tool for paleoecological applications. Journal of Paleolimnology 35: 149–166.

37.

Garcia

Laville

(2001) Importance of floodplain waters for the conservation of chironimid (Diptera) biodiversity in a 6th order section of the Garonne river (France). Annales de Limnologie 37: 35–47.

38.

Giaime

Avnaim-Katav

Morhange

et al . (2016) Evolution of Taman Peninsula’s ancient Bosphorus channels, south-west Russia: Deltaic progradation and Greek colonisation. Journal of Archaeological Science: Reports 5: 327–335.

39.

Giosan

Bokuniewicz

Panin

et al . (1997) Longshore sediment transport pattern along Romanian Danube delta coast. Geo-Eco-Marina 2: 11–23.

40.

Giosan

Donnelly

Constantinescu

et al . (2006) Young Danube delta documents stable Black Sea level since the middle Holocene: Morphodynamic, paleogeographic, and archaeological implications. Geology 34(9): 757–760.

41.

Goiran

Marriner

Morhange

et al . (2005) Evolution géomorphologique de la façade maritime d’Alexandrie (Egypte) au cours des six derniers millénaires. Méditerranée 104: 61–64.

42.

Goiran

Salomon

Mazzini

et al . (2014) Geoarchaeology confirms location of the ancient harbour basin of Ostia (Italy). Journal of Archaeological Science 41: 389–398.

43.

Goodfriend

Stanley

(1999) Rapid strand-plain accretion in the northeastern Nile Delta in the 9th century AD and the demise of the port of Pelusium. Geology 27(2): 147–150.

44.

Gupta

(2007) Large Rivers: Geomorphology and Management. Leeds: School of Geography, University of Leeds and Singapore: Centre of Remote Imaging, Sensing and Processing, National University of Singapore.

45.

Hammer

Harper

DAT

Ryan

(2001) PAST: Paleontological statistics software package for education and data analysis. Palaeontologia Electronica 4: 1–9.

46.

Heiri

Lotter

(2001) Effect of low count sums on quantitative environmental reconstructions: An example using subfossil chironomids. Journal of Paleolimnology 26: 343–350.

47.

Höckmann

(1993) Late Roman Rhine vessels from Mainz, Germany. The International Journal of Nautical Archaeology 22(2): 125–135.

48.

Jones

Schumm

(1999) Causes of avulsion: An overview. Fluvial Sedimentology 6: 169–178.

49.

Juggins

(2007) C2 Version 1.5 User Guide: Software for Ecological and Palaeoecological Data Analysis and Visualisation. Newcastle upon Tyne: Newcastle University Press.

50.

Juggins

(2015) Rioja: Analysis of Quaternary Science Data (Version 0.9-6). R Package. Available at: https://rdrr.io/cran/rioja/.

51.

Kaniewski

Marriner

Morhange

et al . (2018) Holocene evolution of Portus Pisanus, the lost harbour of Pisa. Scientific Reports 8(1): 11625.

52.

Kelterbaum

Brückner

Porotov

et al (2011) Geoarchaeology of Taman Peninsula (Kerch Strait, South-West Russia)–the Example of the Ancient Greek Settlement of Golubitskaya 2. Die Erde 143(3): 235–258.

53.

Klink

(2002) Determineersleutel voor de larven van de in Nederland voorkomende soorten Polypedilum. STOWA conceptuitgave 6: 18.

54.

Klink

Moller Pillot

HKM

(1999) Key to the Higher Taxa and Species of the Lowlands of Northwestern Europe (Expert Center for Taxonomic Identification). Amsterdam: ETI.

55.

McCarney-Castle

Voulgaris

Kettner

et al . (2012) Simulating fluvial fluxes in the Danube watershed: The ‘Little Ice Age’versus modern day. The Holocene 22(1): 91–105.

56.

Mărgineanu-Cârstoiu

Apostol

(2015) La fortification d’Halmyris. Étude architecturale des portes ouest et nord. Caiete ARA 6: 37–78.

57.

Marković

Stevens

Kukla

et al . (2015) Danube loess stratigraphy – Towards a pan-European loess stratigraphic model. Earth-Science Reviews 148: 228–258.

58.

Marriner

(2009) Geoarchaeology of Lebanon’s Ancient Harbours (British Archaeological Reports, S1953). Oxford: Archaeopress.

59.

Marriner

Morhange

(2007) Geoscience of ancient Mediterranean harbours. Earth-Science Reviews 80: 137–194.

60.

Medvedev

Rabinovich

Kulikov

(2016) Tides in three enclosed basins: The Baltic, Black, and Caspian seas. Frontiers in Marine Science 3: 46.

61.

Meisch

(2000) Freshwater Ostracoda of Western and Central Europe. Heidelberg: Spektrum Akademischer Verlag.

62.

Micu

Carozza

et al . (2009) Observations sur l’habitat néo-énéolithique dans le Delta du Danube. In: Bodi

(eds.) Miscellanea in Honorem Annos LXV Peragentis Professoris Dan Monah Oblata. Iași: Institute of Archaeology of Iași, Romanian Academy, pp. 317–336.

63.

Moller Pillot

HKM

Buskens

RFM

(1990) De larven der Nederlandse Chironomidae (Diptera) Deel C: Autoekologie en verspreiding. Nederlandse Faunistische Mededelingen 1: 1–85.

64.

Munteanu

(2012) Porturi fluviale romane din provinciile renane și dunărene (secolele I- III p Chr.). Pontica 45: 181–260.

65.

Neretin

Volkov

Böttcher

et al . (2001) A sulfur budget for the Black Sea anoxic zone. Deep-Sea Research, I 48: 2569–2593.

66.

Opreanu

(2003) Some data on the recent ostracod fauna from the continental shelf of the Black Sea in the Crimea and Sinop areas. Geo-Eco-Marina 2003: 9–10.

67.

Opreanu

(2005) Contributions to the knowledge of recent Ostracoda (Crustacea) distribution in the North-Western Black Sea. Biologie Animala 51: 63–70.

68.

Opreanu

(2008) Ostracode relicte Ponto-Caspice în sectorul Românesc al Marii Negre. Geo-Eco-Marina 14: 57–62.

69.

Panin

(2003) The Danube delta. Geomorphology and Holocene evolution: a synthesis. Géomorphologie: Relief, Processus, Environnement 9(4): 247–262.

70.

Panin

Jipa

(2002) Danube river sediment input and its interaction with the north-western Black Sea. Estuarine, Coastal and Shelf Science 54: 551–562.

71.

Panin

Tiron Dutu

Dutu

(2016) The Danube delta: An overview of its Holocene evolution. Méditerranée 126: 37–54.

72.

Peltier

(2004) Global glacial isostasy and the surface of the ice-age earth: the ice-5G (VM2) model and GRACE. Annual Review of Earth and Planetary Sciences 32: 111–149.

73.

Petts

Möller

Roux

(1989) Historical Change of Large Alluvial Rivers: Western Europe. Chichester: John Wiley and Sons, Ltd.

74.

Pfleger

(1993) Guide des coquillages et mollusques (trans. in French by Temmerman

, 1989). Fribourg: Hatier.

75.

Pint

Frenzel

Fuhrmann

et al . (2012) Distribution of Cyprideis torosa (Ostracoda) in Quaternary athalassic sediments in Germany and its application for palaeoecological reconstructions. International Review of Hydrobiology 97(4): 330–355.

76.

Pitassi

(2011) Roman Warships. Suffolk: Boydell Press.

77.

Preoteasa

Vespremeanu-Stroe

Panaiotu

et al . (in press) Neolithic to modern period palaeogeographic transformations in southern Danube delta and their impact on human settlements in the Enisala-Babadag region. Quaternary International. Available at: https://doi.org/10.1016/j.quaint.2018.09.010.

78.

Reimer

Bard

Bayliss

et al . (2013) IntCal13 and Marine13 Radiocarbon Age calibration curves 0-50,000 years cal BP. Radiocarbon 55: 1869–1887.

79.

Revenga

Murray

Abramovitz

et al . (1998) Watersheds of the World: Ecological Value and Vulnerability. Washington, DC: World Resources Institute.

80.

Rey

Faucherre

Virmoux

et al . (2016) Paleoenvironmental reconstruction of the ancient harbors of King Louis IX (Aigues-Mortes, Rhone Delta, France). Journal of Archaeological Science: Reports 9: 505–513.

81.

Salel

Bruneton

Lefèvre

(2016) Ostracods and environmental variability in lagoons and deltas along the north-western Mediterranean coast (Gulf of Lions, France and Ebro delta, Spain). Revue de Micropaléontologie 59(4): 425–444.

82.

Salomon

Goiran

Noirot

et al . (2018) Geoarchaeology of the Roman port-city of Ostia: Fluvio-coastal mobility, urban development and resilience. Earth-Science Reviews 177: 265–283.

83.

Salomon

Goiran

Pannuzi

et al . (2017) Long-term interactions between the Roman City of Ostia and its Paleomeander, Tiber Delta, Italy. Geoarchaeology 32(2): 215–229.

84.

Salomon

Keay

Carayon

et al . (2016) The development and characteristics of ancient harbours – Applying the PADM chart to the case studies of Ostia and Portus. PLoS ONE 11(9): e0162587.

85.

Seeliger

Pint

Frenzel

et al . (2018) Using a multi-proxy approach to detect and date a buried part of the Hellenistic City Wall of Ainos (NW Turkey). Geosciences 8(10): 357.

86.

Soulet

Ménot

Lericolais

et al . (2011) A revised calendar age for the last reconnection of the Black Sea to the global ocean. Quaternary Science Reviews 30(9): 1019–1026.

87.

Stanley

Bernasconi

Jorstad

(2008) Pelusium, an ancient port fortress on Egypt’s Nile Delta coast: Its evolving environmental setting from foundation to demise. Journal of Coastal Research 24(2): 451–462.

88.

Stefan

(1984) The late Roman fortress at Murighiol. An air photographic study. Peuce 9: 297–310.

89.

Stefanescu

(1982) La formation et l’évolution du delta du Danube. Paris: Comité des travaux historiques et scientifiques, Bibliothèque Nationale.

90.

Stock

Knipping

Pint

et al . (2016) Human impact on Holocene sediment dynamics in the Eastern Mediterranean – The example of the Roman harbour of Ephesus. Earth Surface Processes Landforms 41(7): 980–996.

91.

Stock

Pint

Horejs

et al . (2013) In search of the harbours: New evidence of Late Roman and Byzantine harbours of Ephesus. Quaternary International 312: 57–69.

92.

Stouthamer

Cohen

Gouw

(2011) Avulsion and its implications for fluvial-deltaic architecture: Insights from the Holocene Rhine-Meuse delta. SEPM Special Publication 97: 215–232.

93.

Suceveanu

Zahariade

Topoleanu

et al . (2003) Halmyris I. Monografie arheologica. Cluj-Napoca: Editura Nereamia Napocae.

94.

Syvanne

(2015). Military History of Late Rome 284-361. Barnsley: Pen and Sword.

95.

Thienemann

(1942) Larve und systematische Stellung von Neozavrelia luteola Goetgh. Chironomiden aus dem Lunzer Seengebiet II. Archiv für Hydrobiologie 38: 581–585.

96.

Tockner

Uehlinger

Robinson

(2009) Rivers of Europe. San Diego, CA: Edition Academic Press.

97.

Toonen

Kleinhans

Cohen

(2012) Sedimentary architecture of abandoned channel fills. Earth Surface Processes and Landforms 37(4): 459–472.

98.

Torr

(1894) Ancient Ships. Cambridge: Cambridge University Press.

99.

Tunoglu

Gökçen

(1997) Pontian ostracoda of the Sinop area, Black Sea coast of Turkey. Revue De Micropaleontologie 40(4): 347–366.

100.

Vallenduuk

(1999) Key to the Larvae of Glyptotendipes (Diptera, Chironomidae) in Western Europe. Entomologische Berichten-Nederlandshe Entomologische Vereenigung 59: 128–128.

101.

Vallenduuk

Morozova

(2005) Cryptochironomus. An identification key to the larvae and pupal exuviae in Europe. Lauterbornia 55: 1–22.

102.

Van Dinter

Cohen

Hoek

et al . (2017) Late-Holocene lowland fluvial archives and geoarchaeology: Utrecht’s case study of Rhine river abandonment under Roman and Medieval settlement. Quaternary Science Reviews 166: 227–265.

103.

Vespremeanu-Stroe

Preoteasa

Hanganu

et al . (2013) The impact of the Late-Holocene coastal changes on the rise and de-cay of the ancient city of Histria (Southern Danube Delta). Quaternary International 293: 245–256.

104.

Vespremeanu-Stroe

Zainescu

Preoteasa

et al . (2017) Holocene evolution of the Danube delta: An integral reconstruction and a revised chronology. Marine Geology 388: 38–61.

105.

Zahariade

(1986) Vexillation in northern Dobrudja. Dacia 30: 173–176.

106.

Zahariade

(1991) An Early and Late Roman Fort on the Lower Danube Limes: Halmyris (Independenta, Tulcea county, Romania) In: Maxfield

Dobson

(eds) Roman Frontier Studies 1989, Proceedings of the XVth International Congress of Roman Frontier Studies. Exeter: University of Exeter Press, pp. 311–317.

107.

Zahariade

(2012) Managing environmental archaeology: Some fresh thoughts on old subjects–the Halmyris fort. Peuce 10: 39–52.

108.

Zahariade

Alexandrescu

(2011) Greek and Latin Inscriptions from Halmyris. Inscriptions on Stone, Signa, and Instrumenta Found between 1981 and 2010 (Halmyris Series Monographs II, BAR International Series 2261) (ed. Zahariade

). Oxford: Archaeopress.

109.

Zahariade

Phelps

(2002) Halmyris, a settlement and fort near the mouth of the Danube: Interim report. Journal of Roman Archaeology 15: 230–245.

110.

Zahariade

Suceveanu

Opait

et al . (1987) Early and Late Roman Fortification at Independenta, Tulcea county. Dacia 31: 97–106.

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.

0.00 MB

16.33 MB