1During the period that ran between the 12th and the 16 th centuries, the Horn of Africa went through some key transformations which reshaped the region and set the foundations of political and economic dynamics which in some cases endure to this day. These transformations included the arrival and expansion of Islam, the imbrication of the Horn of Africa in the international trade networks which conformed the Indian Ocean trade system, the development of a network of permanent settlements in territories previously occupied only by nomads, and the emergence, consolidation, and expansion of a score of Muslim states which controlled large territories of the Horn and challenged the power of the Christian Ethiopian kingdom of Abyssinia. Although research on the medieval Islamic archaeology of the Horn of Africa has increased over the last two decades (Loiseau – Dorso – Gleize et al. 2021, Fauvelle-Aymar 2011, Insoll 2021a, de Torres Rodríguez 2020, de Torres Rodríguez – Ruibal-González – Franco Fernández et al. 2022), there are still significant gaps in our understanding of how all these processes developed and correlated. Most of the research has focussed on the role of Islam in the different social and political changes that took place during the medieval period, while less attention has been paid to the Muslim states which ruled the region until recently (Chekroun – Hirsch 2020). To address this lack of information, in 2020 the StateHorn project was launched to study the medieval Muslim states of the Horn of Africa, to understand their strategies used to achieve control, the sources of their authority and legitimacy, and the ways these polities managed cultural, social and religious diversity. The final objective is to extract information useful to understand the governance and stability challenges that states face in the region today. The project is based on previous research in Somaliland (de Torres Rodríguez 2022) and currently conducts fieldwork in Ethiopia and Djibouti , where a field project was launched in 2021 to study Handoga , the only medieval site known in the country so far.

2Traditionally, the history of the medieval period in the southern half of the Horn of Africa has been reconstructed through the use of the significant corpus of Arab, Ethiopian and European – mostly Portuguese – written sources. From this remarkable body of written sources, academics such as E. Cerulli (Cerulli 1931, Cerulli 1941), J. S. Trimingham (Trimingham 1965), J. Cuoq (Cuoq 1981) and U. Braukamper (Braukamper 2002) have been able to reconstruct with some detail the history of the medieval Muslim states of the southern Horn of Africa. Despite some gaps in our understanding of the process of arrival, expansion and development of Islam in the region which are only now starting to be clarified (Fauvelle-Aymar 2011), by the 9th–10th centuries a score of small Muslim principalities existed to the south and east of the Ethiopian highlands, the most important being Shoa. By the end of the 13th century, Shoa was conquered by the neighboring state of Ifat, which in 1285 became the dominant kingdom and started a process of unification of the Muslim principalities under the Walasma dynasty. By the beginning of the 14th century, this expansion led to an increasing conflict with the Christian kingdom of Abyssinia. In 1376 a long period of hostilities started which ended in 1415 with the total defeat of the Muslim armies, the killing of the last sultan of Ifat (Trimingham 1965: 74) and the effective end of that state.

3The disappearance of the sultanate of Ifat did not end the conflict between Muslims and Christians in southern Ethiopia. A new sultanate – Barr Saʿd al-Dīn, often referred to as Adal – emerged as the successor of Ifat in the region around Zeila, and by the last decades of the 15th century the new state challenged the Christian armies and defeat them repeatedly (Trimingham 1965: 82). By the mid-16th century, the Muslim leader Ahmed Gragn launched a war of conquest which took the Muslim armies to the shore of the lake Tana, almost erasing the Ethiopian kingdom. However, the death of Ahmed Gragn soon afterwards, the combination of military defeats, the decay of the trade due to the Portuguese blockage of the Red Sea and the invasion of the Galla (Oromo) groups from the south ended with the sultanate of Adal by 1577.

4Despite being in a strategic position for the trading routes that connected the coast and the important town of Zeila with the Danakil depression and the Ethiopian highlands, references to the region during the medieval period are extremely scant, and none of them mention settlements in the region where Handoga is located, despite being a major site whose size parallels that of other important medieval towns in Somaliland or Ethiopia. This lack of information made Handoga and Djibouti a strategic zone to understand the influence and control of the successive Muslim sultanates which ruled this region.

5The medieval site of Handoga is situated to the west of Djibouti, about 13 km from the border with Ethiopia and 10 km to the west of the town of Dikhil (Fig. 1). The site is located in a semidesert plain but close (around 1 km) to the wadi Chekheiti, which was likely the source of water for the medieval inhabitants of the settlement (Fig. 2). To the southeast of the site there is a low outcrop which has traditionally been identified as a part of the settlement, although some of the structures found there are undoubtedly more modern. Handoga lies in a strategic position for the caravan routes that connected the coast of the Red Sea with the interior of the Horn of Africa, being located at the crossroads of two main itineraries, one of them coming from Lake Abhe through the Dakka mountains, the other one using the basin of the Hanlé river (Cauliez – Gutherz 2020: 191). Its position was likely strategic for the caravans which crossed the territory during the medieval period. It is also likely that the town was included in the territory controlled by the Sultanates of Ifat (1285–1415) and Barr Saʿd al-Dīn (1415–1577), which probably ruled over this region although the characteristics of this control haven’t been properly analyzed yet.

6Although Handoga was identified in 1939 by the French traveler Aubert de la Rüe who referred to it as “Gallagota or Handouga” (Cauliez – Gutherz 2020: 191), it wasn’t mentioned again until 1969 (Kern 1969), and only in the 1970s when archaeological investigations took place at the site. In 1974–75, two archaeological campaigns were conducted by Roger Grau – founder of the Group of Prehistoric, Archaeological and Geological Studies (GEPAG in French) excavated five houses and a mosque of different sizes and shapes and a vessel buried in the open air, at the outskirts of the site (Grau 1982). A third campaign took place in 1980 and included some aerial photographs (Ferry – Grau – Bouvier 1981). One of the most interesting features of these campaigns was the identification of a series of vessels buried inside one of the houses (Fig. 3), which were identified as “censer holes” although that interpretation is not fully accepted.

7Although the results of the GEPAG excavations have been published with some detail, several aspects of these interventions remain unclear. Most of the pottery recovered during the excavation was not drawn, there is not a plan of the archaeological sites and the sketches to identify the houses and the test pits excavated are not very accurate. Moreover, some houses which look excavated do not seem to correspond to the published plans, suggesting that some of the later excavations were not published. For the 1980 campaign there is just a simple note (Ferry – Grau – Bouvier 1981). In the CERD facilities there are some materials (archival and archaeological) that will be studied in the following campaigns to incorporate them into the general study of the site. There are also some doubts about the interpretation of some archaeological features discovered during the intervention, and later reports by Roger Joussaume point to the existence of older archaeological phases not detected during the GEPAG excavations (Cauliez – Gutherz 2020: 191).

8In 2007, a survey was conducted at the site by the PSPCA project (Gutherz – Pène – Omar Ismaël 2007). Although brief, the survey was very successful and increased substantially the knowledge of the site. The survey identified the structures previously excavated by the GEPAG, but also new elements such as the presence of stelae on the outskirts of the town (Cauliez – Gutherz 2020: 194). Moreover, the excavation of a house yielded an astonishing number of objects related to smelting or metalworking (Gutherz – Pène – Omar Ismaël 2007: 18–22). In a shallow (just 10 cm) excavation, more than 1,000 fragments of slag, copper ingots and iron plaques (Gutherz – Pène – Omar Ismaël 2007: 20) were documented, totalizing more than 8.3 kg of metal. This amount of metal is strong evidence that the house was likely used as a metallurgic facility, the only one found so far in the Medieval Horn of Africa. It also adds another layer of complexity to the interpretation of Handoga, which in addition to a strategic position in the caravan routes could have also been a production place for iron or metal objects. Despite these important results and the interest of the CERD in developing the medieval archaeology of Djibouti, research in Handoga was stalled until the 2021 campaign, which aims to develop a comprehensive approach to this relevant site.

9The 2021 campaign aimed to make first contact with the archaeological site, its characteristics and layout, as well as the logistics required to conduct an archaeological project at the site. Its goals included the documentation of the most important archaeological features at the site to identify the structures excavated in previous campaigns. Several test pits were also planned, to document the stratigraphy of the site, to evaluate the depth and state of preservation of the archaeological deposits and to collect well-contextualized archaeological materials and organic samples to be radiocarbon dated. Finally, the campaign included some work at the CERD facilities, to get an idea of the archival and archaeological materials from prior campaigns which could be incorporated into the new project.

10The survey and mapping of the site relied on foot surveys, photographic documentation and GPS georeferencing. Given the short period at the site, surveys focussed on defining the extension of the site and its limits, identifying singular structures, areas of scattering of materials and previously excavated structures. The information was checked against satellite images from Google Earth.

11The settlement is roughly organized on a north-south axis that measures around 600 m in length and about 300 in width, for a total extension of about 16 ha (Fig. 4, Fig. 5). It has an irregular distribution defining two main clusters of buildings, the southern one larger in size. Most of the structures are circular or oval and appear densely packed in clusters, often in what look like compounds or walled courtyards. To the south and east, the structures become more disperse, in what looks like the outskirts of the town. Although no clear urban layout has been established, there are empty spaces within the clusters of structures that could correspond to squares, junctions of streets or other public spaces. In one of these empty spaces, an extensive area of refuse dump area has been identified, and it is likely that empty spaces could be occupied by structures built with perishable materials. The survey has also documented a potential graveyard to the south of the site, although the structures are much deteriorated. To the southwest, a nearby low rocky outcrop also contains some structures. This outcrop was investigated in the 1970s by the GEPAG, which excavated a mosque and a house there and named it “the Acropolis” (Grau 1982: 16), although the hill does not have any defensive value. During its survey, the StateHorn team identified the mosque excavated by the GEPAG, but other structures seem modern, and some of them were confirmed by the local staff as being contemporary cattle pens. Handoga lacks any defences – either topographical or human-made –, a feature that it shares with the Somaliland medieval sites where unwalled sites are the norm (de Torres Rodríguez 2020).

12The structures of Handoga share a common set of features but they also have a wide variability of designs, sizes and constructive techniques (Fig. 5, Fig. 6, Fig. 7). At this moment, it is impossible to determine if this variability is explained by the different functions of the buildings or has a chronological meaning. In general, houses are round or oval, measuring about 5–6 m in diameter, built with basaltic blocks of trapezoidal shape which in most of the houses present a flat face to the interior, giving the impression of well-built structures. Most of the walls are made of two lines of blocks, bound with dry mud and with the space between the blocks infilled with small stones (Fig. 7). In general, no more than two or three rows of stones are preserved, and considering the scarcity of collapse levels in the site it is possible that the upper part of the buildings was made of perishable materials. There are some differences in the constructive technique, with some of the buildings being worse constructed, as happens to the wall of Test Pit 3. When found, doors are consistently oriented to the west.

13Although most of the houses are circular or oval, there are some remarkable differences among them. In several cases, these houses appear linked in rows of two or three, with the walls of the different houses interwoven. In other cases, these structures appear clustered around what looks like courtyards, defining compounds with several houses attached to a larger, surrounding wall. Finally, there is a third type of smaller circular structures (about 4 m in diameter) that often appear isolated and which could correspond to animal pens or other domestic functions. Although scarce, there are also some square buildings in Handoga, one of which was tested during the 2021 campaign (Test Pit 2). At this moment is impossible to determine if the different shapes of these buildings correspond to different chronologies or functions. Regarding singular buildings, two mosques have been documented at the site. The first one, located in the southern cluster, is a nomadic mosque of 10 by 7 m, which was marked by a line of stones on the floor (Fig. 8). Its nomadic nature and the fact that it is located in a peripheral position in the site make unlikely that this was the main mosque of the site, maybe corresponding to a later period or being a mosque used by the nomadic populations and caravans which used the site. The second mosque (Fig. 9) was excavated by Grau in the 1970s in the outcrop to the southwest (Grau 1982: 16). Again, its size and position make it unlikely that this was the main mosque of the site. Besides these two buildings, the only one that shows some differences is a square building of about 9 m of side, placed on a platform to the southwest of the site.

14Materials are relatively abundant throughout the site, although they seem to be concentrated in specific areas (Fig. 10, Fig. 11). The main accumulations of materials have been documented to the east of the site, in an area where erosion has affected the archaeological layers and where abundant bones, hand-made pottery and shells have been found. In other areas of the site, archaeological materials are very scarce, except for cowry shells which are ever present in the site. Many of these cowry shells have the back removed, and the fact that these back fragments appear systematically throughout the site in large amounts could point to the presence of workshops where the cowries were prepared before being sewn to clothes or objects, as has been recently documented in the medieval town of Harar (Insoll 2021b). Leaving aside the cowries, imported materials are extremely scarce. During the survey, only three of these objects were found at the site: a wheel-made base of a medium-sized vessel (Fig. 12, 5b), a big stone (carnelian? agatha?) bead (Fig. 12, 6) and a small fragment of green glass (Fig. 12, 7). Bones are relatively common but concentrated in specific areas, such as the dump to the west.

15During the surveys, several structures excavated in previous projects were documented. We have already mentioned the mosque found at the outcrop, and the 2007 excavation was easily identified thanks to our local guide. The excavations of the 1970s have presented a bigger challenge. The published map where the 1974 excavations were situated (Grau 1982) was far from accurate, and the plans of the excavated houses could not be matched with any of the structures seen during the survey. The inverse situation was also documented: several houses throughout the site showed evidence of having been excavated, although no information was provided in the publications and reports available (Fig. 13, 10). In these cases, they will probably correspond to the 1980 campaign, of which we do not have almost any information. In that sense, two archive photographs provided by the CERD of an unidentified and unpublished excavation of a square house also point in this direction. A small test pit (C-1000) conducted during the 2021 campaign confirmed this aspect.

16Despite time limitations, the survey conducted in 2021 in Handoga has been very useful to identify the main characteristics of the site, getting a first glimpse of the main urban features of the site, planning the 2022 campaign, collecting some interesting materials and setting some working hypothesis (Fig. 13). The absence of a drone has prevented an accurate mapping of the town, which will be one of the priorities of the 2022 campaign and which will help to integrate between the data obtained by the StateHorn project with those from the previous interventions in the site.

17Test Pit 1 was conducted in a well-preserved round house located to the southwest of the site, in a densely built area of the settlement (Fig. 14). The eastern part of the house’s wall had a height of four rows – exceptional for Handoga – but the lack of binding materials among the ashlars and the “emptied” aspect of the interior of the structure suggested that it may had been excavated already. To check this, a small test pit was planned in the southern side of the house, measuring 2 m in length in its east-west axis and a maximum of 1.20 cm in the north-south axis (Fig. 15). The southern side didn’t have a regular shape, as it adapted to the circular shape of the wall. The excavation yielded an extremely simple stratigraphy: after the removal of a layer of a very loose, yellow sand (SU 1001) which corresponded to the surface level and had a depth of 20 cm to the south and just 5 cm to the north, a more compacted layer was documented. This layer (SU 1002) had a depth of just 5 cm and consisted of brown sand of small grain. To the southwest, a small accumulation of greyish sand with some charcoals (SU 1003) was identified, with an irregular oval shape of 18 by 12 cm. The small accumulation was excavated, documenting a small irregular pit of 12 cm of depth (SU 1004), probably corresponding to the imprint of a root. The removal of the SU 1002 in all the test pits gave way to another layer of compacted, yellow-light-brown sediment with whitish imprints, which was identified as the sterile, geological layer on which the house was built. Therefore, the initial impression that the house had been previously excavated was confirmed and the test pit was ended at this point. Although the excavation of the house did not yield any archaeological information, it has confirmed an important aspect to be considered in future excavations at the site: the existence of structures excavated by the GEPAG team but never published. That is also the case of three linked structures to the south of Test Pit 1, which had been undoubtedly excavated before, and the search for any data about these interventions (field diaries, photographs, e.g.) will be a priority for future campaigns.

18Test Pit 2 is situated to the southwest of the site, in a wide flat area covered with archaeological remains, ashes and charcoals identified during the survey, cut by several small crevasses caused by torrential water. Abundant materials were found at the natural profiles caused by erosion. The area (Fig. 16) extends for around 2.500 sq. m and probably consisted of a refuse dump for this area of Handoga. Given the quantities of materials documented on the surface of the dump, a small test pit (1 x 1 m) was planned close to a crevasse, in order to 1) understand the stratigraphy and origin of the refuse dump and 2) collect samples of material culture and organic materials, especially charcoals which can be dated by 14C.

19The excavation (Fig. 17, Fig. 18) yielded a very simple stratigraphy. The surface level (SU 2001) consisted of a shallow (2–4 cm deep) layer of yellow sand with small pebbles and archaeological materials. After its removal, a layer of about 10–12 cm was documented, consisting of very loose light brown sand mixed with ashes of a greyish color (SU 2002). Charcoals are very abundant, along with animal bones, pottery sherds and some pieces of slag. Although during the excavation no successive layers of ashes were documented, some packs of ashes were visible in the profile of the excavation, corresponding to recurrent episodes of deposition of trash. After the removal of the SU 2002, a small (5 cm) layer of very compacted yellow sand was documented, which has been interpreted as the original floor of the area which has been hardened by the deposition of hot ashes and charcoals after the clearing of hearths or other fires. This layer was not removed, but the clearing of the profile of the crevasse showed that under this yellow stratum the geological level of this area was located, consisting of an orange-brownish compacted sand.

20As could be expected in a refuse dump, materials were abundant considering the small size and depth of the trench: 32 sherds of hand-made pottery were collected (9 of them providing some information about the shape, decoration or function of the vessel), along with 27 shells or shell parts of different species. Ten lithic pieces and seven iron slags were also collected, along with a relevant number of animal bones. In general, all the materials recovered were very fragmented and eroded. Given the characteristics of the deposits, the most plausible interpretation is that this large refuse dump was generated by the clearing of the hearths of the nearby houses, and maybe, by the nearby foundry area excavated in 2007, which is just 30 m to the south. The presence of some slag fragments found in the test pit could be an indication of this use. The dating of charcoals collected at the SU 2002 will provide further information to determine the chronology of this dump, and in following campaigns a more selective dating could be made to detect the dates of the beginning and end of the use of this dump. Additionally, the big amounts of charcoals found at the test pit will allow us to determine some other interesting aspects, such as the types of wood selected for fuel.

21Test Pit 3 is located to the north of the site, close to the ravines that run into the Wadi Chekheiti. It is located in one of the few square structures located in Handoga, a house 5.20 by 5.40 m, east-west oriented and with a total area of 28 sq. m. The house shares its north wall with a second, round structure (Fig. 19). The house was selected for a number of reasons: to check if the stratigraphy, chronology and material culture documented in this structure were similar or different from those circular or oval; to excavate a test pit in a previously unexplored area of the settlement, and to define the potential structures found in the building (doors, thresholds, partitions, etc.). The final objective of this test pit was to determine if differences in shape and location within the site could correspond to a different chronology. The test pit was oriented east-west, measuring 1.5 by 1 m, and was situated at the northwest corner of the building, covering the area where a gap in the wall pointed to the existence of a door (Fig. 20, Fig. 21, Fig. 22).

22The test pit offered a very simple stratigraphy. After the removal of the surface layer, made of very loose yellow sand (SU 3001) which had just 3–4 cm of depth, a new layer of about 20 cm of depth was found, made of brown, coarse sand mixed with gravel (SU 3002). The removal of this deposit led to the identification of the door’s opening, measuring 120 cm and without any kind of doorjambs or threshold. It also showed that under the SU 3002 a sterile, geological level was discovered, identical to the one found at Test Pit 2 (Fig. 18), which marked the end of the test pit, although 10 extra cm were excavated to check that there were no lower rows of the wall. The walls were very poorly preserved (just two rows of height), consisting of irregular rectangular blocks of volcanic stones, presenting a flat surface to the interior (Fig. 23). No foundations of the walls were found during the excavation.

23Test Pit 3 has confirmed the scarcity of the archaeological deposits at Handoga which was already noted by the French archaeologists in the 1970s. Unfortunately, the excavation has not provided much information about other aspects of the buildings, such as living floors, hearths or other internal structures. Archaeological materials are very scarce, too. Only one fragment of bone and a piece of a dark blue glass bangle have been recovered during the excavation (Fig. 12, 8). The bangle has close parallels with pieces found in Somaliland and Ethiopia, offering a chronology between the 15th and 17th centuries. Their provenance is unclear, but places of production are known to exist in Egypt, Syria, Iran and India (González-Ruibal – de Torres Rodríguez – Franco Fernández et al. 2021: 15). It’s one of the few imported objects found in the 2021 campaign, and in general in all the research in the Handoga, where imported materials are very scarce. The results of the test pit, unfortunately, have been so poor that it is not clear if the rest of the building will be excavated in future campaigns.

24The last test pit excavated in 2021 was situated at the eastern part of Handoga, close to the track that runs through the site and about 110 m to the south of Test Pit 3. It is located close to several structures and mounds eroded by torrential rains, forming shallow crevasses which run into the wadi Chekheiti. It is an area where several scatterings of materials were found during the survey, including the stone bead which appeared just 60 m away. The test pit was located in a large compound defined by a wall of 70–80 cm wide, which seems to surround an irregular circular area of 15 m in diameter and an area of about 200 sq. m. The southern side is not visible, but there are some hints that other smaller structures could be attached to it. The wall is made of big volcanic stones, but it has poorer quality than those of the structures of Test Pit 3, visible in the more irregular aspect of the faces of the wall. The objective of the test pit was to explore one of these compounds documented throughout the site, which could correspond to courtyards or animal pens. The test pit was set on the eastern side of the compound, in an area where the wall was well preserved and it was estimated that the depth of the sediments would be greater. The test pit was set on a NE-SW axis, measuring 1.5 by 1 m (Fig. 24).

25After the removal of the loose yellow sand that corresponded to the surface level (SU 4001), mixed with several stones corresponding to the collapse of the wall, a layer of dark brown, compacted sand was documented throughout the test pit. This layer (SU 4002) of 22 cm of thickness did not provide any archaeological materials, but underneath it an accumulation of ashes, charcoals and some small bones was detected extending throughout the center of the test pit (SU 4003) (Fig. 25, Fig. 26). In the western profile and at the same level, a fragment of profusely decorated pottery was documented. The pottery presented radical differences from the rest of the sherds found during the survey and the excavations, in terms of firing, shape and decorative patterns (Fig. 12, 4a–b). Once the SU 4003 layer was removed, two new aspects of the test pit became evident. The first one was that the wall had a single row of stones and after the removal of the SU 4002 the walls did not go any lower. Therefore, SU 4003 belonged to an older occupation stage in this area. This impression was confirmed by the discovery, at the southern side of the test pit, of part of a circular pit that continued under the unexcavated area. This pit was under the layer of ashes (SU 4003), and therefore it also belonged to an older phase than the walls.

26The interest of this discovery led to an 0.5 m extension of the test pit to the south (for a total of 1.5 by 1.5 m), to document and excavate this pit (Fig. 27, Fig. 28, Fig. 29, Fig. 30, Fig. 31). Once the SU 4001 and 4002 SU units were removed (the layer of ashes coded as SU 4003 did not appear in this extension), the outline of the pit, which had 60 cm of diameter, was defined and the whole test pit was photographed and documented with photogrammetry. The pit had a diameter of 65 cm, 26 cm of depth and slightly sloping walls. It was infilled by a single layer of dark brown coarse sand (SU 4004), very homogeneous, with some small fragments of bones and charcoals. Once reached the bottom of the pit, the negative structure was named SU 4005 and the sterile geological level documented in the other test pits was also found here (Fig. 28).

27The proper identification of an older phase in Handoga – suggested several times but never properly described – is an important step in the process of establishing a well-defined chronological framework for the site. To achieve this, two radiocarbon samples were dated, the first one coming from the level of ashes (SU 4003) and the second one from the sediment infilling the small pit at the center of the test pit (SU 4004). The results (Fig. 32, Fig. 33) were extremely coherent, offering an almost identical chronological range: 1221–1273 AD for HA21-0021 and 1220–1274 AD for HA21-002 (both ranges with a 95,4 % of probability). Therefore, it seems that this older phase could be confidently established around the mid-13th century – at least, in this part of the site.

28Archaeological materials were present but not especially abundant, consisting mostly of small fragments of undiagnostic pottery, cowries and bones (Fig. 34). The only relevant pieces were the aforementioned black pottery fragment with a complex decorative pattern of incisions and punctuations, and a black sherd with a polished surface which resembles materials from the Ethiopian Highlands (see Fig. 12, 3a–b, Fig. 12, 4a–b).

29Although brief, the excavations conducted at Handoga during 2021 have provided an important set of data for the understanding of the site, which will be of immense help in the following campaigns. The first of these data has been the confirmation of the scant depth of the preserved archaeological structures and deposits. None of the test pits excavated in 2021 had more than 40 cm of depth, and the maximum height of the buildings was three rows of stones. That small thickness had already been noted by the teams excavating in Handoga, and greatly influenced the excavation methodologies applied to the site: especial care must be given to prevent the over-excavation of structures and the documentation of the different archaeological layers, and a more extensive approach can be taken when working at the site. In that sense, the identification of the geological levels at different parts of the test pits will help us to define the moment the excavation process must be stopped.

30A second important data is the scarcity of archaeological materials, confirmed both during the survey and the excavation. Most of these materials were found extremely fragmented and eroded, suggesting that those could be left behind during what seems a systematic and gradual abandonment of the site. That scarcity contrasts with the data from previous excavations, which gathered an important set of well-preserved, in some cases almost complete vessels, not to speak of the huge amounts of metal products found at the 2007 excavation. These differences will have to be explained during the next campaigns, when larger areas are excavated at the site.

31Finally, a third outcome of the test pits excavations is the confirmation of the existence of older occupations in Handoga, something which had been suggested before but had never been archaeologically documented. The excavation of Test Pit 3 has shown clearly that at least two phases of occupation existed at the site, although it is unclear if they were consecutive, if there was a hiatus among them and if they can be documented throughout the site. The radiocarbon dating of the charcoal samples collected at the bottom of the pit will soon provide a better understanding of the chronology and duration of this first occupation phase of Handoga.

32Although limited in time and extension, the surveys and excavations conducted in 2021 have allowed a preliminary overview of the distribution, types, characteristics and chronologies of the archaeological materials found in Handoga. This overview can be checked against the results of previous excavations, although these materials have not been fully published. In general, archaeological materials are not very abundant throughout the site, something relatively surprising considering the size of the settlement. Materials appear concentrated in certain areas, in some cases related to erosive processes that have dismantled deposits, such as the refuse area where Test Pit 2 was dug. In these cases, pottery and cowries are the most abundant materials, along with bones, charcoals and in some cases small fragments of slag. One of these accumulations corresponded almost completely to pierced cowries and their removed upper parts, suggesting that an area of Handoga could have been dedicated to the preparation of cowries for their attachment to clothes or objects. Imported materials were extremely scarce throughout the site: only one small fragment of green glass, a stone bead and a wheel-made fragment of pottery were collected during the survey.

33The observations made during the survey were confirmed to some extent during the excavation of the four test pits. The test pits have yielded a limited sample of materials. Test Pit 2 – despite being a refuse dump – yielded just 32 fragments of pottery and 27 of cowries and other shells, with some pieces of iron slag, lithics and some bones. In Test Pit 3, only a fragment of bone and another fragment of a glass bangle were found during the excavation, while Test Pit 4 yielded a more abundant but equally limited sample of materials, consisting of pottery sherds (14), shell fragments (35) and a relatively abundant sample of bones. In general, the objects found during the excavations were eroded, fragmented and decontextualized: no in situ materials were found during the excavations.

34This situation contrasts with the information provided by previous researchers in the settlement, who described the presence of materials in the site as abundant and found several pieces in situ in the houses excavated in the 1970s and 1980s. The excavation conducted in 2007, significantly, found an astonishing amount of slag fragments and metal objects during their excavation of a house, suggesting that the scarcity of materials found in 2021 could be due to differences in the use or abandonment of spaces throughout the settlement. What seems constant throughout time is the low percentage of imported materials at the site: during the 1980 campaign, for example, just six imported pieces were found (Ferry – Grau – Bouvier 1981: 58); and no imported pottery has been found so far in the site, contrasting with what happens in most of the medieval sites in the region, where imported pottery from Arabia, Middle East and Asia are the most common objects.

35The overwhelming majority (99.4 %) of pottery collected in Handoga during the 2021 campaign corresponds to hand-made pottery of medium quality, corresponding to vessels of medium size (Fig. 12, 1a–k). Firing seems to have been conducted in an oxidizing environment, although in some cases black or grey stains in the pottery walls point to an open-air firing which would provide a mixed result for the pieces. Walls correspond to two types: a coarse type with middle-sized temper particles of volcanic origin, and a more refined one with more regular, smooth surfaces, although there is no evidence of any treatment slip or polish techniques. The sole exception is a fragment of pottery found in Test Pit 4 (Fig. 12, 3a) which presented a burnishing treatment similar to those found in the Ethiopian highlands (de Torres Rodríguez 2017: 232). Temper particles are mostly of volcanic provenance, something logical considering the geological environment, although white limestone intrusions can also be found in most of the fragments.

36The fragments found so far correspond to medium-sized vessels of globular or semi-globular shape, with several variations in the orientation and shape of the rims. Most of the pottery has inward rims which are usually round (Fig. 35, 6, 10–12, 16–17) although in some other cases the border has a flat shape (Fig. 35, 8, 14, 18). Another type consists of vessels with straight necks but slightly everted rims (Fig. 35, 2–3). Finally, two different shapes have been documented in the sample: one semi-globular bowl of big size, straight walls and flat, beveled rim with a “T” profile (Fig. 35, 13), and the base of a small bowl which will likely correspond to a censer (Fig. 35, 5). Besides this base, no other hand-made bases have been found so far in the sample. Two types of handles have been documented: a small round handle with a circular section which seems to have been attached vertically to the vessel (Fig. 12, 1b), and a horizontal type with a circular or oval section, located at the upper part of the body (Fig. 35, 15, 17). The position of these handles in the upper part of the vessels suggests that they were used to put ropes through them to hang or transport the vessels. Knobs have also been documented, with both triangular and oval shapes, and again in the upper part of the vessels (Fig. 34, 16, 18). The decoration is very scarce and limited to incisions. Only two fragments had these decorations, one of them consisting of parallel horizontal lines (Fig. 35, 8), the other having diagonal series of parallel incised lines (Fig. 35, 18; Fig. 12, 4a). As happened with the handle or knobs decoration patterns are placed in the upper part of the vessel, close to the neck.

37The sample gathered during the 2021 survey is similar to the examples published in the 1970s reports, which documented more complete vessels, mostly globular with inward-flat or everted rims (Grau 1982: plate 3, 1, plate 2). Small and circular (Grau 1982: plate 3, 1) or horizontal handles (Grau 1982: plate 3, 2) were also found, along with the base of a censer identical to the one found in 2021 (Grau 1982: plate 4, 1). Decorative patterns are more varied though, including fingernail impressions, X-like incisions on the upper part of flat rims or punctuations defining bands of triangles (Grau 1982: plate 2, 1–7). Although a bigger sample is required, the pottery recovered during the 2021 season is also very similar to the materials found in medieval settlements in Somaliland, both in terms of shapes and decorations, including very characteristic shapes like the flat rims with incised decorations (Fig. 12, 2), globular vessels with marked inward rims and the “T” profiles. Handoga is just 150 km from a large cluster of medieval sites in Somaliland and therefore is likely that there were cultural relationships between these two areas.

38Only three fragments of pottery stand out from the sample. The first one is the aforementioned small varnished fragment of black pottery, a type of wall treatment characteristic of the Ethiopian highlands (de Torres Rodríguez 2017: 232). The second one (Fig. 12, 4a–b) is a much-decorated sherd fired in a reducing environment, with a long straight neck and an oval shape. The piece is decorated with two series of parallel horizontal incised lines close to the rim, and a band of three horizontal lines at the shoulder, the middle one made of circular punctuations and the external ones incised. The body of the vessel is decorated with triangular bands made of parallel diagonal incised lines combined with punctuated parallel series. The shape, wall thickness, firing and decoration of this pottery are completely exceptional – although a sherd found in the 1975 excavation shows a similar decorative pattern (Grau 1982: plate 2, 8). Given that this pottery was found at the lower occupation level of Test Pit 4, it could correspond to an older period of the settlement, which according to the 14C samples analyzed in Test Pit 4 would be dated in the mid-13th century. The foreign origin of this pottery seems very evident and could be related to influences coming from the Ethiopian highlands. The last fragment that stands out of the sample is a wheel-made base (Fig. 12, 5b), which so far is the only one of this type found in Handoga and which was obviously imported.

39As has been mentioned before, imported objects are very scarce in Handoga, something coincident with the situation in many medieval sites in Somaliland, but something difficult to explain if we assume the strategic position of Handoga defended by the prior excavator of the site. During the 2021 campaign, only four objects of clear foreign origin were found at the site: the aforementioned wheel-made fragment of pottery, a carnelian bead found during the survey (Fig. 12, 6), a tiny fragment of green glass also found during the survey (Fig. 12, 8) and part of a dark glass bangle (Fig. 12, 7). Surprisingly, the most common imported objects found in most of the medieval sites in this part of the Horn of Africa (Chinese, Persian or Arabian wares) are completely absent in Handoga.

40This scarcity of imported materials is coherent with the results of the previous campaigns in the site, which have yielded very few objects, too. In the 1975–1976 excavations only several fragments of glass vessels, bangles and beads were found (Grau 1982: 8). The drawing of one of these bangles (Grau 1982: plate 4, 4) presents a decorative pattern very similar to those found in Somaliland and dated in the 15th–16th century. In 1980 the materials were a bit more abundant: a stone bead identical to the one found in 2021, several glass beads and a bangle and a couple of pendants made of mother-of-pearl (Ferry – Grau – Bouvier 1981: 58). In the report of the brief 2007 survey no imported materials are described, but there is a reference to two Arab coins found in the place of which no further information is provided (Cauliez – Gutherz 2020: 192).

41As was already noted during the excavations in the 1970s (Grau 1982: 8), lithic materials are relatively abundant throughout the site, either obsidian or flint flakes or debitage products from the knapping process (Fig. 12, 9a–c). In one case a retouched edge was identified. Three polished stone objects were also found during the surveys. This type of object was profusely documented in prior campaigns, showing a great variability of types including stone balls and grindstones (Ferry – Grau – Bouvier 1981: 58). No mills or grindstones were documented in 2021, although they undoubtedly were used throughout the site. Other materials include iron slag pieces of small size, especially abundant in Test Pit 2 – the refuse dump close to the foundry structure excavated in 2007, where these slag fragments were very abundant (Fig. 12, 10a–b). Finally, bones were relatively abundant both during the surveys and the excavation. Most of the bones collected were small and fragmented (Fig. 36), corresponding to long bones although the species are still to be determined.

42One of the most remarkable characteristics of Handoga is the presence of several scatterings of shells throughout the site usually in flat open areas (Fig. 37). They consist mostly of cowries although other bigger types of shells have also been documented. Cowries, which probably represent 99 % of the shells, appear in either whole or with the back removed which is the usual process in order to attach them to clothes or objects. What is more interesting is that along with the cowries, the removed dorsa of the shells have also been documented in great numbers, pointing that the process of preparation of the cowries took place in Handoga (Fig. 12, 11). A similar situation has been detected in the site of Harla (Ethiopia), where the identification of these fragments has led to the proposal of a marine shell workshop (Insoll 2021b). Given the number of dorsa found at the site, that could also be the case with Handoga.

43The preliminary excavation at Handoga and the analysis of the published materials from previous excavations have provided an interesting set of data that allow to establish a basic framework for the interpretation of the site, as well as some elements of comparison and discussion within the wider, regional context. If we consider the urban layout of the site, Handoga shares some similarities with nearby Somaliland sites of the same chronology such as Amud, Abasa or Hasadile (de Torres Rodríguez 2020, de Torres Rodríguez – Ruibal-González – Franco Fernández et al. 2018): unwalled settlements with a disperse layout, made of scattered buildings with empty spaces which would likely correspond to courtyards and squares. This similarity has some differences, too: buildings in the medieval towns of Somaliland are systematically square or rectangular (de Torres Rodríguez – Ruibal-González – Franco Fernández et al. 2020), while in Handoga all the structures except three or four are circular or oval. In fact, although houses appear grouped in clusters both in Djibouti and Somaliland, the Handoga ones seem to define big walled compounds to which round houses are attached, while in the few documented cases in Somaliland such as in Fardowsa the buildings are the ones that define the space, the walls just closing the spaces between the houses. In that sense, both the circular houses and the compounds suggest a strong nomadic substrate in the layout of Handoga, something cleverly detected by its early excavators who defined it as a “city of nomads”. This nomadic character is also reinforced by the structures found at the cemetery, located at the south of Handoga and which presents important differences with those found in the urban settlements of Somaliland. In general, they consist of accumulations of stones surrounded by rings of stones (Fig. 38), a type of burial closer to nomadic traditions and radically different from the cists found in Somaliland graveyards.

44Finally, another radical difference (at least, as of today) is the absence of an urban mosque in Handoga as happens in all the Somaliland sites. None of the teams working at the site have been able to identify one of these buildings. Two mosques have been found in Handoga, one of them a nomadic-style mosque situated to the south of the site and another, small one at the outcrop which is in a very peripherical position with regards to the settlement. There are at least three possible explanations for the absence of urban mosques in Handoga. The first one would be that they have not been located yet, something unlikely considering the good visibility of archaeological structures at the site and the time different teams have spent in the site. A second explanation is that the nomadic mosque found to the south was actually the mosque of Handoga, which would reinforce the idea of a “petrified nomadic camp” that the clustering layout of the houses and the type of graves at the cemetery suggest. Finally, it could happen that the population at Handoga was actually non-Muslim, and therefore no mosques were necessary – the nomadic one a later addition to the site. Only future research in the area can solve this problem. Of course, the fact that Handoga has a nomadic-like layout doesn’t imply it was inhabited temporarily. The documentation of a foundry or metal workshop indicates that the site was occupied on permanent bases, and if we attend to the information at Test Pit 4, for a long time.

45Regarding the chronology of the place, most of the information points to a 14th–16th centuries period for the last occupation moment of Handoga. Two radiocarbon dates have been obtained at Handoga, offering a 16th century and a 14th–15th range, although the exact results and contexts are unknown (Cauliez – Gutherz 2020: 192). These chronologies are coherent with some of the materials found during the different excavations, especially glass bangles (Grau 1982: 8; Ferry – Grau – Bouvier 1981: 58) which are dated between the 14th and 16th centuries in Somaliland (González-Ruibal – de Torres Rodríguez – Franco Fernández et al. 2021: 15). Considering this still scant information, it seems that the final stages of Handoga are contemporary to the medieval sites of Somaliland, where similar archaeological materials have been located. There are also several pieces of evidence that point to a much older occupation of Handoga: two Arab coins dated in the 11th and 13th centuries are cited in several publications (Cauliez – Gutherz 2020: 192) although no further information is provided. The only other reference to older chronologies comes from parallels of the polyped vessels found during the 1970s excavations, which could have parallels in the Harar area dating from the 8th to the 12th centuries (Gutherz – Pène – Omar Ismaël 2007: 18). The sum of this data would offer a preliminary range of the 11th to the 16th centuries, and will be completed and improved by a series of radiocarbon dates that will be made during 2022. Regarding the specific chronology for the end of the site, it seems very likely that, as has been already pointed out, the abandonment of Handoga was progressive and non-violent, maybe due to environmental changes which made the area unfit for a permanent settlement (Grau 1982: 6, 16).

46The sum of all these data raises some interesting issues to be discussed in future campaigns. The first one is related to the identity of the inhabitants of Handoga. It has been assumed that the inhabitants of Handoga were Muslim, but the lack of an urban mosque, the style of the burials and the presence of non-Muslim architectural features such as a stele found in the outskirts of the settlement could imply that the population of Handoga was non-Muslim or only slightly Islamised. Considering that Handoga is located at the core of the sultanates of Ifat and Adal, this aspect would add an interesting perspective on the relationships between Muslims and non-Muslims in a confessional state.

The article presents the results of the first campaign carried out at the medieval site of Handoga (Djibouti) by the StateHorn project, based at the Institute of Heritage Sciences of the Spanish National Research Council. The aim of the campaign was to assess the site's potential in order to launch a long-term project focusing on the study of the town's urbanism and way of life. The campaign included a systematic survey of the site and the excavation of four test pits, which revealed evidence for two archaeological phases at Handoga. The results of the campaign suggest that Handoga was an important urban centre on the medieval routes linking the Gulf of Tadjoura with the interior of Africa, of which very little is known.