Introduction

1The Farasān Islands were brought to the attention of the academic world, interested in the ancient Red Sea, through the publication of two Latin texts, written locally by members of military detachments from two separate Roman legions – one stationed in Egypt (at Nicopolis) and the other in the province of Arabia (at Bosra).1 Surveys2 reported built remains which evoke traditions known in the interior of the Arabian Peninsula; pottery similar to traditions of the Late Bronze Age and Early South Arabian period attested on the coastline facing the archipelago, and two Nabataean sherds on the site of Ġurrayn. The surveys3 carried out in the wake of these discoveries have made it possible to record many sites from various periods, to revisit previously known sites and, in particular, ancient sites whose chronology ranges from the period of expansion of the kingdom of Saba to Late Antiquity. Substantial remains, dated from the first centuries CE by surface material and by the archaeological excavations4 carried out since 2013 in the site numbered Wādī Matar 2, illustrate the Roman presence on the archipelago.

2The southwestern coast of Saudi Arabia has yielded notable sites dated to earlier periods (from the Late Bronze Age to the Ancient South Arabian period, i.e. the first half of the 1st millennium BCE), to Late Antiquity, and to the Early Islamic period, but no remains are currently firmly attested in the region of Jizān, facing the Farasān Islands in the coastal area, during the first centuries CE.5 Thus, we find parallels in material cultures between Farasān and the mainland in the early 1st millennium BCE and in the Early Islamic era.6 Regarding the first centuries CE on the other hand, there is a difference between the two territories. At this time, Farasān appears to stand out as an “anomaly” in the regional archaeological landscape, due to the presence of Roman soldiers confined on the archipelago. The material from Farasān mirrors the maritime trade on the Red Sea and the Roman presence, whereas this type of material (Nabataean fine ware, transport amphorae, Roman architectural blocks) seems, up until now, to be totally absent from the coastline of the Jizān region.7 It is therefore undoubtedly the insular nature of Farasān that justified Rome’s choice to send a military detachment there. Without speculating on the status of the islands prior to the Roman presence, of which we know virtually nothing,8 we can nevertheless observe that Rome benefited from the weakness of the regional powers (Ḥimyar and Adulis) to settle there.9 It also implies a knowledge of the region’s advantages on the Roman side, which is not otherwise known from classical sources.

3Recent surveys in the Farasān archipelago have made it possible to draw up a cartography of the places occupied during the first centuries CE that reveals a strategic organization of this insular territory by a foreign group, the Roman military, taking advantage of the resources the place had to offer.10

4We will first refer to the ancient sources that mention the region and testify to a certain knowledge of this area, while apparently omitting the Farasān archipelago. This will allow us to underline the specific characteristics of Farasān in the region. We will then closely examine three contemporary sites dated to the first centuries CE which present different but complementary characteristics.

Sources on the southwestern Arabian Peninsula during the first centuries CE

5Classical literature offers little more than identification based on toponymy while archaeology only yielded limited material of uncertain date in the coastal area of southwestern Southern Arabia.

Archaeological finds

6Previous research yielded isolated findings that indicate human settlement in the coastal area of the southern half of the Red Sea during Antiquity (fig. 1). A coin hoard of 181 silver Sabaean coins (of “imitation ancient Athenian tetradrachm” type) dated to the late 4th century BCE was found in Bājil at the foot of the ‘Asīr mountain east of al‑Ḥudayda.11 According to the work of B. Davidde, this find indicates the importance of the route from the highlands down to the Red Sea at this period, through the Wādī Sihām and leading to the site of al‑Hāmid (dated to the first half of the 1st millennium BCE).12 Contemporary sites, however, remain absent from the documentation so far. As for the first centuries CE, fragments of bifid handles of Dressel 2/4 amphorae were found south of al‑Maḫā’ in Yemen13 and in the city of as‑Sāwa during Yemeni excavations, a site linked to the coast.14

Fig. 1 – Map of the southwestern coast of the Arabian Peninsula with location of mentioned sites in the text and ancient toponymy as described in classical sources (S. Marion de Procé 2020).

7Around Jizān however, no solid evidence was recovered. Ancient South Arabian inscriptions are mentioned in the hinterland of Jizān, not far from the site “al‑Rayyān” and in the foothills of Wādī Ḍamad, but with too few details to allow dating.15 Sites may have been obliterated by later occupations or may have been destroyed16 either by the urban development or by the environmental conditions. Given the limited evidence, however, it is not possible to describe the nature of human occupation in the 1st century CE. The settlements evidenced around Jizān rather seem ascribed to early 1st millennium BCE and Early Islamic contexts. In terms of foreign presence, the Latin inscriptions from Farasān and the absence of any foreign pottery on the coastal area facing the islands support the idea that the Romans were exclusively settled on the archipelago at that time.17

On the historical geography of the region

8The apparent absence of archaeological sites dating to the end of the 1st millennium BCE and the first centuries CE is even harder to understand as the classical authors do mention villages and people living on the coastal plain. These are never referred to as trading hubs but, even so, one would expect to find surface remains somewhere – again, further surveys and study of local pottery productions are needed to clarify this aspect.

9The most descriptive text of the Red Sea ports and settlements is the handbook written in the mid 1st century CE, the Periplus Maris Erythraei. Unfortunately, the coast of Arabia from Leuke Kome to the “burnt island” (Katakekaumenè, which could be identified either with the Jabāl al‑Ṭā’īr18 or with the Ḥanīš volcanic islands)19 is only mentioned in a short paragraph which states that several tribes inhabit the coast and speak various languages, and that the region should be avoided given the threat of piracy and dangerous reefs.20 This text stresses the danger of piracy in the south-eastern part of the Red Sea in the 1st century CE. Previous publications21 contain extensive information on the region’s toponymy found in the texts of Strabo (Geography, 15‑20 CE, relying on earlier Hellenistic and 1st century BCE sources), Pliny (Natural history, third quarter of the 1st century CE also relying on earlier sources, themselves relying on Hellenistic sources) and Ptolemy (Geography, mid 2nd century CE, probably using a source from the reign of Trajan).22 These texts mostly reproduce geographical and ethnological information collected from earlier sources that allow identification with modern toponyms or provide hints as to where one should look for these ancient settlements.

10In several texts, we find ethnonyms that can be linked to the area of Jizān in various forms. In Pliny’s text, the Mamaean coast with its goldmines23 is mentioned before the “Canauna district and the tribes Apitami and Casani” (litus Mamaeum ubi auri metalla, regio Canauna, gentes Apitami, Casani),24 the latter can clearly be seen to reflect the name of Jazān.25 Similarly, the ethnonym “Kassanitai” found in Ptolemy’s Geography,26 and possibly the “Kasandroi” mentioned in Agatharchides,27 are likely to refer to the ancient communities living in the region of Jazān at this time.

11Pliny’s “Pteros”28 has been proposed as a possible deformation of the original name or a hellenized name of Farasān evoking the wings of the numerous birds of the archipelago (which Ptolemy’s “Hierakôn” could echo, as hierakes could refer to falcons and hawks or to some carnivorous marine animal, both found in Farasān).29 Regarding Pliny’s “Sambrachate”,30 a location in the area of Jizān is argued by C.J. Robin in this volume based on South Arabian epigraphical sources.31

12Diodorus of Sicily offers a description faithful to the environment at this period and mentions another form of the ethnonym of the people of Jazān. The author describes the difference between the land of the Debae in whose territory runs a river carrying gold and, south of it, the land of the Alilaei and Gasandi, who also find gold in galleries. He writes: “it is not fiery hot, like the neighbouring countries, but is often overspread by mild and thick clouds, from which heavy showers and timely storms that make the summer season temperate”.32 Similarly, the British Red Sea and Gulf of Aden pilot33 notes how the landscape changes north and south of the port city of Jizān. There is no doubt that “Gasandi” mentioned here is the same as “Casani/Kassanitai/Kasandroi” and that these occupy an area covering the coastal plain as well as at least part of the highlands where gold is mined.

13Theophrastus’ “Mamali”,34 an incense-bearing region, and Pliny’s “litus Mamaeum” may be linked to the same location,35 possibly facing al‑Shuqaiq where a couple of reefs are named “Mamali” in the Red Sea and Gulf of Aden pilot36 some 50 km north of Farasān.

14As for Ptolemy’s “Badeo basileion”,37 a couple of names are possibly linked to the ancient location in the hinterland of Jizān: al‑Baiḏ and al‑Badi. The mention of a “basileion” (understood as a “royal palace” in this context, possibly a town where a local sovereign resided) in the region is noteworthy and reflects the possible presence of a political power there. Ptolemy’s mention of “Ambe” refers to a location in the region of Jazān, possibly Jizān itself, located in the “Kassanite” country.38 We have no idea if the Farasān Islands were part of the ethnic groups mentioned in the classical sources as being present in the region of Jazān.

15It is therefore surprising that so few vestiges from Antiquity are known while the ancient sources testify to the presence of small local coastal settlements. Several reasons can explain this: these settlements were built in perishable material,39 now lost; there is no pottery study that we can rely on for identification since a lack of surveys has prevented the sites from being identified. As said above, no pottery that can be easily ascribed to the first centuries CE has been identified in the publications. Lastly, no text is known of so far in the Jazān region, except for Farasān, where South Arabian and Latin inscriptions are documented.40 Only some unpublished ones are known of and, here again, further archaeological and epigraphic surveys will undoubtedly shed more light on the map of human settlements in the coastal plain in the first centuries CE.

Farasān in the first centuries CE

16The first centuries CE undoubtedly represent a period of revitalization of the Farasān archipelago after centuries of apparent abandonment,41 a reactivation of the economy that is easily understandable in the context of the development of commercial routes in the Red Sea and the Roman presence on the islands. The latter is attested by two Latin inscriptions: the first one was discovered in a modern cemetery east of Farasān city where ancient carved blocks were reused and the second was found in the village of al‑Quṣār, whose name possibly reflects the presence of a Roman monumental building. The modern city of Farasān did not yield any other significant remains. The building mentioned in the first Latin inscription has probably been dismantled and the stones identified in the cemetery may come from the area (either from Farasān, where an ancient site would have been covered by modern buildings, from the natural hill of Qal‘at Farasān, north of the city, on which now stands an Ottoman fort or from Jabal al‑Kniseh, bearing an explicit toponym reflecting the archipelago’s Christian past, and where ancient sherds and remains have been spotted).42 It is less likely that ancient blocks were brought back from the oasis of al‑Quṣār, but the option cannot be discarded as al‑Quṣār was the summer resort of the residents of Farasān until the former was abandoned around 1975. Surveys carried out in recent years have produced a non-exhaustive map of the occupation of the islands during Antiquity and more specifically during the first three centuries CE. The eastern part of the main island has been the subject of more systematic surveys which have provided a more complete view of the distribution of sites in this area (fig. 2).

Fig. 2 – Map of the Farasān Islands with location of sites mentioned in the text (S. Marion de Procé 2020).

17In the field, the architecture of carefully cut blocks and the presence of ceramic or numismatic material allow us to date the sites. These consist of terra sigillata sherds, fine Nabataean ceramics, Roman transport amphorae (Dressel 2/4, especially Campanian), South Arabian transport amphorae and Himyarite silver coins that provide the most solid elements for dating a site by surface material.

18In addition to the sites about to be presented, we should mention the sites of Qurayyāt, located on the island of Farasān al‑Kubra, and Sagīd, on the island of Sagīd, which offered visual landmark and took advantage of some of the best anchorages that the archipelago has to offer. The ashlar masonry and surface material allow us to date them from Antiquity (first centuries CE). The inventory of other sites on the rest of the archipelago has to be completed by further surveys.

Al-Quṣār

19The village of al‑Quṣār is the place where the second Latin inscription comes from. The original size of the latter must have reached 59 cm in height and 90 cm in length, although only the bottom right corner of it is preserved.43 The possibility of identifying the original building, or any other elevation, into which it was inserted in at al‑Quṣār is almost nil. The architectural blocks preserved in Mr I. Miftah and Mr I. Sayyādi’s private museums suggest the total dismantling of Roman buildings. However, some foundations still remain visible on the surface whereas many carved blocks were used to erect the walls of modern houses (fig. 3). While the heart of the ancient site has been destroyed, information gathered during our surveys gives us a general overview of this site (fig. 4). A hectare of dense vegetation in the heart of the oasis could still conceal remains, but the area is impenetrable for the time being. A large square building was surveyed in January 2020 to the west of the area covered with vegetation, widening the original site’s boundaries. At present, we can only rely on the observation of reused blocks in the modern walls to assess the type of masonry used in the site.44 Two blocks (half of a “so-called Nabataean” capital and column carved as a monolith in the local coral limestone and a miniature naiskos imitating very similar examples discovered in Marina al‑Alamein/Nicopolis in Egypt), found in al‑Quṣār by I. Miftah and I. Sayyādi are directly linked to the presence of the vexillatio of the IInd Legion Traiana Fortis sent from the province of Egypt and their original camp at Nicopolis.45 One of the many enigmas still unsolved is the presence of arched stelae (fig. 5) whose closest parallels we found in Baḥrein and which come from funerary sites of the Tylos period.46 The material collected from the surface consists mainly of modern pottery and glass. However, Roman pottery is visible in areas close to concentrations of reused carved ashlars (amphora Dressel 2/4, fragments of Roman rooftiles locally made with crushed shell temper).

Fig. 3 ‑ Modern building in al‑Quṣār exhibiting reused carved blocks (S. Marion de Procé/MIFA 2019).

Fig. 4 – Location of remains mentioned in the text in the oasis of al‑Quṣār (satellite image courtesy of Google Earth Pro; G. Davtian, S. Marion de Procé/MIFA 2020).

Fig. 5 – Example of arched stela from al‑Quṣār (S. Marion de Procé/MIFA 2018).

20The oasis of al‑Quṣār provides fresh water drawn from wells dug in the bedrock. The area has a fertile clayey soil still used today for the cultivation of date palms and sorghum. Irrigation of these crops, similar to the ones we see today, was by rainfall alone. A study conducted in January 2019, and based on the observation of current agricultural practices by C. Cosandey,47 established that seeds were planted a few days after one of the heavy winter rains between December and January.

21In order to erect their monuments, the occupants of al‑Quṣār had stone in abundance, the archipelago being made up of coral limestone, outcropping everywhere. Apart from faint traces of surface quarrying, an area to the east of the site has been identified as an ancient quarry (fig. 6) of yellow-coloured limestone, characteristic of the area (in contrast to Wādī Matar where the coral limestone presents a greyer colour for example). While most of the carved stone objects discovered so far have an exogenous style, which could indicate that craftsmen accompanied the Roman garrison,48 it should be noted that contemporary Ancient South Arabian inscriptions were discovered on this site and suggest a local component of the site’s population (the castrenses?),49 that may well be the remains of slightly earlier or later occupations. Other seemingly contemporary South Arabian inscriptions are known in the archipelago (one comes from the modern village of Ḫutūb [Khutūb] on the island of Sagīd, while others are of unknown provenance).

Fig. 6 – Quarry east of al‑Quṣār, large blocks in the course of extraction are still visible (F. Villeneuve 2011).

22As for the pottery production, the presence in the soil of large bright red clay areas and the availability of various temper agents (vegetation, sand, crushed shells) offered the resources to produce their own wares. A kiln, erected in fired bricks bound with lime, was recorded in January 2019 (fig. 7). The absence of material in the surrounding area, with the exception of what is probably an Islamic coin, makes it impossible to date it. This same absence of material leads one to think that it was a lime kiln rather than a pottery workshop or an oven for cooking food. Finally, two pit tombs (fig. 8) covered by stone slabs have been identified in the south-eastern part of the oasis. Both have unfortunately been cleared out.

Fig. 7 – Remains of a kiln in the south-eastern limit of al-Quṣār (S. Marion de Procé/MIFA 2019).

Fig. 8 – Pit tomb carved in the bedrock and a cover slab still in place (F. Villeneuve 2011).

23Some 700 m east of the oasis, one finds the undated fort of Qal‘at Luqmān, built with finely cut blocks. Easily visible from al‑Quṣār, it offered a clear view of the eastern shores of Farasān al‑Kubra and the sheltered bay of Tubta (fig. 9). A visibility study produced by G. Davtian50 allows the assessment of the scope of this observation tower (fig. 10). A large part of the island is visible. The area located to the northwest of the modern city of Farasān is slightly obstructed by small natural reliefs. Villeneuve’s hypothesis of a dismantled Roman fort in the area of Farasān fits well here as a fort in that area would have offered an intermediate outpost between Qal‘at Luqmān and the northwest of modern Farasān city. In addition, a fort located there would have offered a good view for monitoring the two neighbouring bays.

Fig. 9 – Qal‘at Luqmān remains (east of al‑Quṣār) with views to the west towards the oasis and to the east towards the coast (S. Marion de Procé/MIFA 2014).

Fig. 10 – Visibility study from the sites of Ġurrayn and Qal‘at Luqmān (data: G. Davtian; CAD: S. Marion de Procé/MIFA 2019).

24The oasis of al‑Quṣār offers many advantages and is undoubtedly one of the sites chosen by the Romans to settle. Its location on a low area only protected by the fort of Qal‘at Luqmān makes us think that it was likely to have been part of a wider network of sites that had complementary roles.

Ġurrayn

25Ġurrayn is located on a rocky outcrop northwest of the modern town of Farasān and east of the sheltered Harīd bay. A visibility study was also carried out from the point of view Ġurrayn offers (see fig. 10). At a man’s height on the site’s natural relief, one can already see the hills of Wādī Shāmi. At a height elevated to 8 m, from one of the site’s buildings, visibility is extended to other sites such as al‑Qurayyāt in the centre of the main island. Visibility to the southeast, on the other hand, is limited by the natural reliefs where the modern city of Farasān is built. This supports the hypothesis mentioned above that a fort was erected in the area to complete a surveillance network developed by the Romans.

26The site has been known since the first surveys in the 1980s and was visited by W. Facey and B. Rihani shortly before.51 The monumental architecture of the buildings, of which only two to three courses remain, bears witness to the importance of this site, which extends 350 m north-south and 150 m east-west (fig. 11). Successive visits by different teams have revealed two wells, one of which is old, dried up and bears an Ancient South Arabian inscription on the curbstone. The second well is now used by farmers to irrigate the neighbouring field. The site is surrounded by large silty areas, clearly visible on satellite images, providing fertile soil whose cultivation certainly dates back to Antiquity. A necropolis southeast of the site includes a few tombs of uncertain date, possibly dating back to the Late Antique or Early Islamic period (at least two tombs were marked with a stone engraved with a cross). One tomb52 is also visible at the southern entrance of the fenced site, south of the remains. It was looted in 2018 and the collapse of the structure revealed several courses of perfectly cut blocks. This structure is reminiscent of the Quṣār tombs described earlier as well as of the masonry of a looted tomb in Wādī Shāmi (of Late Antique date).53

Fig. 11 – Aerial orthorectified view and top plan of the remains in Ġurrayn and view of the remains to the southeast (orthophoto: G. Davtian 2019; photo: S. Marion de Procé/MIFA 2019).

27Four non-successive phases of occupation can be deduced from the remains. The earliest period of occupation, attested by the surface pottery material alone and possibly by the South Arabian inscription on the well, dates back to the first half of the 1st millennium BCE. The pottery material tentatively ascribed to the earliest phase is concentrated on the eastern flank of the plateau. This concentration could be explained by later clearing of the Ancient South Arabian levels. The large monumental constructions probably date from Antiquity, judging by the material visible on the surface of several of these structures (Nabataean sherds in ashy layers). The third phase attested in Ġurrayn led to architectural modifications and can be dated to Late Antiquity (presence of engraved crosses and Ayla/Aqaba amphorae as well as possible Aksumite sherds). The last phase is clearly located in the south-eastern part of the plateau. The remains consist of small architectural units made of small unhewn stones, whose elevation still reaches about 2 m. The dating proposed by R. Carter based on the observation of sherds is 15th‑16th century.54

28Without any excavation, one can only presume, from the material and architecture, that this site was occupied by Romans in the 2nd century CE. Given the monumental architecture and its prominent location overlooking the plains and the sheltered Harīd bay, it could have fulfilled an important administrative role.

Wādī Matar

29The third and last settlement to be addressed is located in Wādī Matar. This vast clay plain south of the village of al‑Muharraq is dotted with several historical archaeological sites, most of which date back to the Ancient South Arabian period and to the second half of the 2nd millennium BCE (fig. 12). Many areas have a clayey nature that allows dry agriculture, still practiced today in small scattered fields that are the only human occupation of wādī, now dedicated to the protection of local gazelles. Ploughing is carried out before the heavy annual rains and sowing is done a few days later, without any further water supply being necessary to grow melons and watermelons. The agricultural potential of this microregion is obvious and has certainly motivated the settlement of human groups from the earliest times. The other factor that probably presided over the choice of this area is the presence of underground fresh water. The extent of this water table cannot be assessed at the present time, but the presence of a vast area of vegetation, to the south of which an ancient well is located, bears witness to its importance.

Fig. 12 – Schematic top plan of the remains in Wādī Matar 2 and view to the southeast of the fenced archaeological area (S. Marion de Procé/MIFA 2019).

30The Late Bronze and Early Iron Age periods are documented by several structures, some of which bear Ancient South Arabian inscriptions (first half of the 1st millennium BCE), and by the ceramics found during surveys and excavations carried out in Wādī Matar 2B located west of the “dolina” characterised by the presence of thorny shrubs of the acacia type and capers. The observation of the built remains, obviously modified several times, and of the archaeological finds (Dressel 2/4 amphorae, terra sigillata, Nabataean unpainted fine ware) suggest that the site was reoccupied in the 1st to 3rd centuries, at the very time of the Roman presence a few kilometres to the north.

31¹⁴C dating of the upper archaeological levels in two sectors (WM2A and WM2B) indicated that the last occupation in these places was indeed dated between the middle of the 1st and the end of the 2nd century CE (1895+/-28 BP, WM2A, wood); and the second sample between the middle of the 1st century BCE and the middle of the 2nd century CE (1967+/-38 BP, WM2B, wood).55 These samples, analysed by C. Bouchaud,56 which had an ashen appearance, contained numerous archaeo-ichthyological (fish vertebrae and jaws) and malacological (in large quantities, crushed, giving the sample its ashen colour) remains, mirroring the significant presence of Murex-type shells on the surface. In view of the quantities, the production of purple-dye on the site seems to be excluded. These shells are however sometimes used as fuel, or as a temper agent for the production of ceramics, due to the robustness of the pottery produced in this fashion.57 The samples contained coals vitrified by exposure to very high temperatures. The species represented in these test samples and used as fuel are local species characteristic of desert environments (Balsamum and acacia wood), still well represented in the area. These preliminary observations thus seem to indicate a self-sufficient economy taking advantage of the immediate environmental setting (plain of Wādī Matar and coastline less than 3 km away).

32The WM2A area comprises about ten small structures, the majority of which are single-celled. The concentrations of material from one structure to another seem to vary and suggest that this zone, isolated from the other structures, could be dedicated to production (presence of metallic slag, forge).58 The excavations carried out in January 2020 by S. Bert Geith and P.‑M. Blanc59 on building (3), which was thought to be a forge, do not invalidate this hypothesis. The shallowness of the archaeological deposits resting on the outcropping bedrock does not provide solid confirmation as ancient levels have been destroyed by subsequent occupations. The settlement in sectors WM2A and WM2B is contemporary with the reoccupation of area WM2C, a sacred enclosure whose original date could date to the middle of the 2nd millennium BCE (fig. 12).60 The excavations carried out in 2019 and 2020 and placed under the responsibility of F. Villeneuve61 and S. Mazurek62 have made it possible to stratigraphically confirm this reoccupation after centuries of abandonment. The central building as we see it today was built in the early centuries CE, probably in the 2nd century, in the middle of older structures. Previous soundings by C.S. Phillips and the author had provided material that suggested such a scenario; the weak stratigraphy around the access stairs, however, had not made it possible to establish the preexistence of Ancient South Arabian levels in the sequence. The material discovered during excavations contains a lot of spolia from the earlier levels, the sediments present having been used for the backfill of the temple. There are numerous fragments of easily recognizable pottery (South Arabian storage and transport jars, Dressel 2/4 amphorae, Nabataean fine ware63) and a Ḥimyarite coin dated to the reign of ‘Amdān Yuhaqbid (80‑100 CE). The local common pottery is difficult to identify at the moment due to the lack of references in the region and the absence of well preserved levels so far. Few cult furnishings have survived, but we should mention here the discovery of a finely carved bull’s head made of calcite comparable to South Arabian examples from the first centuries CE as well as a small copper-alloy bell ornamented with theatre masks.64

A network of sites

33These three ancient sites show that the Roman presence was expressed through the erection of different facilities, established to control the islands and the sea surrounding them. This presence led to a reoccupation of previously abandoned sites (such as Wādī Matar). High sites (Qal‘at Luqmān, Ġurrayn) are favoured for watching over the neighbouring sea, especially the sheltered bays of Hadar and Tubta; this fits in well with the hypothesis that the Roman presence was motivated and justified by the need to stem the piracy that raged along the Red Sea coast of the Arabian Peninsula. The sites are provided with fresh water, arable land and building materials. Evidence, still incomplete at this stage but nevertheless existing, points to local ceramic production (for making common ware, storage jars and also construction material: Roman roof tiles, fired bricks). Local civilian communities obviously benefited from the Roman presence; entailing a reactivation of the economic activity and drawing the commercial routes to Farasān. Civilian communities are represented by contemporary South Arabian texts and by ceramic productions which have yet to be identified. The presence of Mediterranean pottery and straw-tempered South Arabian transport jars, all found around the Red Sea and the Indian Ocean as far as India,65 indicates that the archipelago was indeed a stopover on the commercial networks.

34The reconstruction of a temple66 after several centuries is interesting in several respects. Although we do not know whether the deity originally worshipped, ‘Aṯtar ḏū-Rayd(ān?),67 is the same in the 2nd century CE, it is remarkable that the place has retained its sacral function despite the centuries that separate the two phases. Given the mixed nature of the material and the rectangular top plan that could be ascribed to either Roman or South Arabian culture68, it is difficult to say who frequented this temple in this phase. The presence of imported material in the levels of occupation nevertheless underlines the exchanges between the people of al‑Quṣār and the people of Wādī Matar. Craft activities, fishing, and the collection of shells were probably the domain of local people. The shell middens located on the coast illustrate the exploitation of marine resources. These structures are characteristic of prehistoric times, when the subsistence economy was largely based on malacological resources. A recent study, however, mentions three middens located to the west of the bay of Janaba, dated by ¹⁴C between 450 BCE and 350 CE, containing a large proportion of fish bones and marine mammals in addition to shellfish.69

35The observation of this network opens up new perspectives in the search for port and coastal infrastructures, which will guide research by combining the observation of the distribution of land sites and the environment of the archipelago’s bays. In Wādī Matar up to the Early Iron Age, it has been suggested that the coastline must have been further north inland from where it is today, due in particular to the silting up of palaeobays such as the one located south of Wādī Matar, probably protected by a dune barrier.70 During the Roman period, it seems that this part of the coast was no longer in use due to silting, reflecting the notable absence, apart from Wādī Matar 2, of sites occupied in the early centuries CE. The fort of Qal‘at Luqmān associated with the site of al‑Quṣār makes it possible to monitor the eastern channel of the archipelago and the bay of Tubta. The fact that a small modern port and a military settlement are still located there speaks for the qualities of the place, both strategic and geographical. However, in the absence of archaeological evidence, the presence of an ancient port there remains a hypothesis, and the fort may well have been an observation post for the area facing the mainland. Other sheltered bays, such as Harīd and Sagīd, close to which ancient sites have been recorded, may also have been anchorage places for a fleet. Pending a better definition of the ceramic production and solid dating of the sites in question, these remarks remain hypotheses.

Conclusion

36The Farasān Islands were an obvious defensive asset in the southern Red Sea aimed at protecting the commercial routes, not relying on one small military settlement only but rather on a network of strategically located sites. The islands were chosen specifically for their insular nature and the natural protection such a context offers. The reasons behind the choice of Farasān are multiple. A first one is its size and location: it is easily located and offers a large land area. Despite the shallow waters and numerous low islets and coral reefs around the archipelago, the wind regime favours it rather than its African counterpart, the Dahlak, when sailing up and down the Red Sea. It also offers a channel to reach the mainland and terrestrial routes.71 Secondly, the sources are lengthy about the dangers of the southern half of the coast of Arabia: a military presence was much needed in the eastern part of the basin at a time when the regional powers could not handle Red Sea matters as they were dealing with internal struggles. Lastly, this military enterprise was planned as a long-term presence and Farasān offered all the resources a large community needed: fresh water, vegetation, cattle, stone, clay, fish, maritime fauna and shells as well as sheltered bays to station a fleet, and a civilian community of South Arabian culture that welcomed a foreign presence. The abandonment of the military system however must have occurred sometime in the 3rd century CE as no later material is found associated to the Roman pottery scatters in al‑Quṣār and levels in Wādī Matar 2.

37If we are to compare this configuration to other Roman settlements on islands, we can gain some insights from the examples of the Balearic Islands in the western Mediterranean Sea. Mallorca and Menorca were conquered by Rome in 123 BCE as a response to the piracy that developed in the region.72 In Mallorca, the location selected for founding the city Pollentia (Alcudia) gave a clear view of the two large adjacent bays. In Menorca, the port of Sanisera (Sanitja) was chosen by Rome for settling a military camp due to its sheltered geographical setting. A military castellum in the 4th century CE was settled on the island of Formentera in a strategic location in order to control the trade routes and to offer a clear view of the island and the coast; a watchtower was built on a natural relief to extend the visibility over the surroundings.73 In both cases, however not contemporaneously, we see the Roman community taking over the geography of the territory to take advantage of its strategic assets and the development of civilian settlement, living off trading and exchanges with the foreign occupants. These examples provide welcome, and otherwise lacking, comparisons with the Roman strategy applied to occupy insular territories outside the boundaries of the Empire. In Farasān, the territory seems to have been well assessed prior to the settlement of the garrison. They had knowledge of the islands’ assets and settled where they deemed it was most strategic: in high places offering clear views of protected bays and the sea from where an incursion may have occurred, in an oasis and in sites well provided with water and arable lands. In both contexts, the western Mediterranean and the southern Red Sea, the local communities do not seem to have put up any resistance. On the contrary, they appear to have thrived following Rome’s choice to send military detachments there. Island archaeology has much to teach us about the specific needs and the strategies adopted by Rome in various regions. Broader comparisons may help to interpret Farasān’s Roman settlement history, as no other external sources mention this episode which was cut short by the crisis the Empire experienced.