Al-Bakhra (Avatha), from the Tetrarchic Fort to the Umayyad Castle

Transcription

1 LEVANT Pp Al-Bakhra (Avatha), from the Tetrarhi Fort to the Umayyad Castle Denis Genequand Counil for British Researh in the Levant, P.O.Box 519, Jubaiha, Amman, Jordan This artile presents the results of arhaeologial and arhitetural survey and small-sale exavations at al-bakhra, a late Roman and early Islami site south of Palmyra. The site began as a Tetrarhi fort attributed to an auxilliary avalry unit. During the Umayyad period the fort was modified and a huge extension delimited by a new enlosure wall was added. There are also numerous houses, funerary strutures and hydrauli devies. The Umayyad astle or qasr belonged to one of the Companions of the Prophet and his desendants. It was also the loation where the aliph al-walid b. Yazid was killed shortly before the ollapse of the Umayyad dynasty. For this reason al-bakhra is known in some detail from the written soures and it is of great interest to ompare the information from these soures with the arhaeologial evidene. 1. Introdution The site of al-bakhra, now also known loally as al- Bkhara or al-bukhayra, is situated 21 km. to the south of Palmyra in Syria. Well-known amongst islamiists for having been a possession of Nu man b. Bashir and the plae where the aliph al-walid b. Yazid was murdered, it has however never attrated muh attention from arhaeologists. For the latter, the site was mainly a late Roman fort briefly desribed at the beginning of the twentieth entury by Musil and Wiegand. As part of a Syrian-Swiss projet studying the Umayyad settlements in the Syrian steppe, and working under the aegis of the Swiss-Liehtenstein Foundation for Arhaeologial Researh Abroad (SLFA, Zürih) and the Diretorate General of Antiquities and Museums of Syria, the site was surveyed in June 2002 and soundings were made in July 2003 (Genequand 2003). This artile presents the general results of the survey and the site s history. A more detailed report on the arhaeologial results will be published in the near future. The first arhaeologial reports on the site were produed by Musil (1928, ), followed shortly by Wiegand (1932, 13). Both of these sholars onentrated on the site s main struture, the Roman fort, to the exlusion of all other features, and provided plans giving an approximate idea of its layout. The site was briefly mentioned by Poidebard (1934, 52, 59, 66 67) and was again surveyed during the late 1980s by Bauzou, who provided a more aurate desription of the site in an unpublished PhD thesis (Bauzou 1989). It was also Bauzou who definitively identified the anient name of al- Bakhra through the study of the milestones on the Strata Dioletiana (Bauzou 1993). 2. Historial soures Al-Bakhra is seurely identified with the antique Avatha, aput viae of a setion of the Strata Dioletiana, the Roman road built or rebuilt under Dioletian and running in a south-west to north-east diretion aross Palmyrena (Bauzou 1993, 34 35, 46 47). The name Avatha is also known from the Notitia Dignitatum, whih plaes a avalry unit there at around AD 400: the Equites promoti indigenae (Notitia Dignitatum, Or. XXXII, 22). It is also known that the same avalry unit was already garrisoned there under the Tetrarhy, as onfirmed by a Latin military insription dated AD (Bauzou 1993, 47 49). This insription is engraved on a stele urrently reused as a pillar for an hypostyle building most probably a mosque situated near the southern orner of the Roman fort. This Late Antique oupation is also doumented by another Latin and two Greek insriptions disovered during the reently undertaken work, details of whih will be published in a forthoming artile. There are no written soures for the Byzantine period, but the site reappears in early Islami soures under its present name, al-bakhra. The 225

2 226 LEVANT most detailed aount is found in al-tabari (Ta rikh, II, ). This aount relates how the aliph al-walid b. Yazid fled to al-bakhra at the time of Yazid b. al-walid s rebellion and how he was killed there on AH 27 Jumada II 126/AD 15 April 744. Al- Bakhra was the astle (qasr) of Nu man b. Bashir, a Companion of the Prophet and of the first Umayyad aliphs, and it still belonged to his desendants (Banu Nu man b. Bashir). It is listed amongst several other fortified sites, inluding Tadmur/Palmyra and the very lose site (1.6 km.) of al-hazim/al- Sukkariyya, where the aliph may have hosen to flee the rebellion. As none of these sites orresponded to a plae where al-walid was keen to go, he finally deided on al-bakhra. Al-Tabari gives two onseutive aounts of the events that happened after al-walid left al-aghdaf (identified with Qasr al-tuba in Jordan), until he was killed in al-bakhra by a fore led by Abd al Aziz b. al-hajjaj b. Abd al-malik. While mainly onentrating on the events, al-tabari s text nevertheless provides many details about the site. Its main feature was a astle, equally alled qasr or hisn in Arabi, presented as well-fortified and built by the ajam. This latter term usually means Persian but, as demonstrated by Kennedy (2001, and 192, n. 6), it also has alternative meanings and in this ontext would be better interpreted as Roman. A village (qariya) is also desribed, where al-walid s troops were supposed to find fodder for their horses. The third omponent of the site was the plae where al-walid was staying at first, whih is desribed as a fustat, that is a tent or a amp. In both aounts, the aliph, feeling that the danger was beoming greater, retreated to the astle/qasr as a last resort. Aording to the first aount, al- Walid exhanged some words with the assaillants from the top of the astle s entrane and, after having been insulted, went inside one of the rooms to read the Koran. The assaillants then limbed the rampart and went into the room and deapitated al- Walid. The seond aount differs slightly and suggests that al-walid left his tent or amp to go inside the astle. Surprisingly, the entrane to the astle was only barred by a hain under whih Abd al- Aziz s troops easily entered, with only one man having to limb the astle wall. They killed al-walid with their swords and then ut off his head. Several other soures refer to al-bakhra, but in less detail. Ibn Qutayba (Ma arif, 266, 186), in a paragraph devoted to al-walid b. Yazid, only states that the latter was killed by Yazid b. al-walid in al- Bakhra. Al-Masudi (Muruj, vol. 4, 49) mentions that al-walid b. Yazid was killed in al-bakhra, desribing the site as a village (qariya) in the area of Damasus where he was also buried. The same author indiates in another work (Tanbih, 324) that al-walid b. Yazid was at the time of his death in a astle (hisn) alled al-bakhra. Al-Bakri (Mu j. am, 141) states that there are two plaes alled al-bakhra, one on the Persian Gulf and a seond in Bilad al-sham, whih is desribed here. The site in Bilad al-sham was alled al- Bakhra due to the bad smell of the plae. Al-Bakhra effetively means the stinking. Indeed, Yaqut (Mu j. am, I, 523, s.v.) explains that the name of al-bakhra means something that smells bad, stagnant water for example, and ites a text narrating al-walid b. Yazid s last moment. The aliph, having learned of the arrival of his enemies, asked for a Koran in order to be killed like his unle Uthman. He was then killed and parts of his body ut off: his head was sent to Damasus in a basket, whilst one of his hands was taken by a dog. The site seems to have been no longer important in al-bakri s time. In fat, al-bakri does not give any indiation that it was still settled. Yakut may well have meant a plae of whih only the name survived, beause it was the plae of al-walid b. Yazid s murder. There is no mention of al-bakhra later in the medieval period. 3. The site The site, now ompletely ruined and uninhabited, is loated on a slight prominene in a vast plain. The entre of the plain is formed by the Sabkhat al-muh, a depression following the southern slopes of the Palmyrenian range and where the rainwater olleted in the wadis gathers to form a seasonal lake. The presene of an artesian spring in al-bakhra has been the main reason for human settlement at the site, inluding the Roman military oupation. The spring, at the bottom of a deep hole in the bed-rok, is situated in the immediate viinity of the southern orner tower of the Roman fort. There are numerous flint artefats from the lower and middle Palaeolithi, in partiular some very fine handaxes, in the sediments deposited by the spring and sattered around it. The site measures approximately 600 m. from west to east and 750 m. from north to south, overing around 40 hetares (Fig. 1). The Roman fort, with a huge extension to the north-east, oupies the entre of the site. The northern part of the site is mainly formed by a dense network of enlosures, while the south-western, southern and eastern parts onsist of numerous houses and a few funerary monuments and strutures. To the south of the fort s

3 DENIS GENEQUAND Al-Bakhra (Avatha), from the Tetrarhi Fort to the Umayyad Castle 227 BAKHRA Plan du site Fondation Suisse-Liehtenstein pour la Reherhe Arhéologique Relevé et dessin: Sophie Reynard, Christian de Reynier Murs maçonnés observés Murs maçonnés restitués Puits et anaux N Courbes de niveau: équidistane = 0,5m m Figure 1. General plan of al-bakhra (drawing: S. Reynard and C. de Reynier).

4 228 LEVANT BAKHRA Plan du hâteau Fondation Suisse-Liehtenstein pour la Reherhe Arhéologique Relevé et dessin: Sophie Reynard, Christian de Reynier Murs maçonnés observés Murs maçonnés restitués Puits et anaux Courbes de niveau: équidistane = 0,5m m N Figure 2. Al-Bakhra : plan of the Tetrarhi fort and the Umayyad extension (drawing: S. Reynard and C. de Reynier). Figure 3. Al-Bakhra : view of the Tetrarhi fort from the south (photo: D. Genequand). gateway, near the southern orner tower and opposite the spring, are the remains of a hurh and a mosque. Two middens overed by sherds are situated to the east and north-west of the fort. Six wells dug into the bedrok provide an additional water supply to the spring A short distane away to the east and to the southwest from the site are two large neropolis that have been almost ompletely robbed out. To the northeast and south-east are two smaller units onsisting of a few rooms and enlosures, as well as some smaller middens easily distinguishable as ashy deposits with a high onentration of surfae sherds. Surfae sherds olleted systematially aross the different parts of the site show an oupational sequene extending from the seond to the ninth enturies AD.

5 N DENIS GENEQUAND Al-Bakhra (Avatha), from the Tetrarhi Fort to the Umayyad Castle 229 BAKHRA Restitution du amp romain Fondation Suisse-Liehtenstein pour la Reherhe Arhéologique Dessin: Christian de Reynier m Figure 4. Al-Bakhra : partial reonstrution of the original plan of the Tetrarhi fort (drawing: C. de Reynier). 4. The Tetrarhi fort The first phase of the Roman fort onsists of a retangular struture measuring m. (exterior dimensions) (14,972 m 2 ), with a south-east to north-west orientation (Figs 2, 3 and 4). The main gateway of the fort is loated on the south-east side of the building. The struture has four fanshaped orner towers,. 10 m. in diameter, and seven U-shaped interval towers whih projet. 10 m. Two of these U-shaped towers flank the gateway in the entre of the short south-east side of the fort (Fig. 5), two others are situated at regular intervals on eah of the long sides of the fort (Fig. 6). There is only one U-shaped tower in the entre of the short north-west side of the fort, situated on the same alignment as the gateway. The seond U-shaped tower visible on this side, whih is smaller and irregularly plaed, is a later addition. This is also the ase in the transformation of one of the south-western towers into a retangular plan, as well as the addition of another retangular tower immediately beside it. The rampart wall is 3 m. thik, where fully preserved, and is entirely onstruted of mediumsized roughly ut bloks. The gateway and the towers are built of larger and better dressed stones. A sounding was done inside the fort against the rampart, between the gateway and the south-eastern

6 230 LEVANT Figure 5. Al-Bakhra : the gateway flanked by two U-shaped towers and the interior of the fort (photo: D. Genequand). Figure 6. Al-Bakhra : north-east rampart of the Tetrarhi fort with the U-shaped towers (photo: D. Genequand).

7 DENIS GENEQUAND Al-Bakhra (Avatha), from the Tetrarhi Fort to the Umayyad Castle 231 Figure 7. Al-Bakhra : view of interior of the Tetrarhi fort from the north-eastern rampart (photo: D. Genequand). Figure 8. Al-Bakhra : door of the Prinipia in line with the gateway of the fort in the bakground (photo: D. Genequand).

8 232 LEVANT tower. The rampart was preserved up to 3.70 m. in height here, inluding a m. high foundation. Four main oupation layers were reognized. The lowest one, at the same level as the foundations, preeded the onstrution of the fort and was dated to the seond or third enturies AD. The seond layer orresponded to the fort s original oupation and was dated to the late Roman period. It was overed by a fill, on the top of whih a rough paving had been laid. This paving formed the third oupation layer. This was followed by a very thik (1.50 to 1.80 m.) rubbish dump made of organi sediments and numerous sherds. These sherds were dated from the fourth to the eighth/early ninth enturies AD. The last and most reent oupation layer was only 0.30 to 0.40 m. below the surfae. It onsisted of a rough, partially paved floor and a number of installations related to some kind of grinding ativity. No material was found to provide a date for this layer, but it appears to be quite late (seventeenth to nineteenth enturies?). The small number of sherds retrieved from the layers under the first floor relating to the rampart provide a rather indeterminate terminus postquem of the seond/third enturies AD, a date whih nevertheless orresponds with the muh more preise one given by the epigraphi evidene (AD ). In addition, it will be shown below that this Tetrarhi date for the fort s onstrution is also onfirmed by its arhiteture. Inside the fort, the strutural remains are largely overed by sedimentary deposits, but their state of preservation is generally good. This is amply demonstrated by the fat that several arhes are still standing, but hardly emerge from the ground, as well as by numerous doors, indiated by vertial monolithi posts, of whih only the lintels are missing (Fig. 7). Only a few walls are visible on the ground however. The reason for this may be that they were mainly built in mud-brik or with smallsized bloks that deay more rapidly. It appears that rows of rooms were organised parallel to the rampart and were separated by narrow streets. It is impossible without exavation however, to assign a preise date to most of the strutures visible inside the fort. Nevertheless, some of them appear more learly to belong to the original onstrution, while others seem more likely to relate to a later period. Some of the door posts, for example, may belong to the first military onstrution. This is ertainly the ase for a very wide doorway situated near the entre of the fort that faes, and is exatly in line with, the gateway (Fig. 8). This doorway probably represents the entrane to the Prinipia. In ontrast, a number of elements of portioes built with re-used materials (olumn shafts, orinthian, ioni and dori apitals) as well as an industrial setor omprising in situ oilpress uprights, ertainly post-date the military use of the fort. It is also worth noting here that all the very finely arved orinthian apitals observed by Musil (1928, 142 and fig. 39) at the beginning of the twentieth entury, and again by Bauzou (1989, 342) in the 1980s, have now disappeared. The few remaining ones are very badly preserved. Conerning the later phase of the Roman fort, the rampart shows several modifiations to its original state, whih may be related to later oupations, at a time when it had lost its stritly military role, but was keeping a defensive funtion. Firstly, there were some alterations to the rampart masonry and the replaement of one of the U-shaped towers on the south-western side of the struture by two retangular ones. This, as well as the opening of a small entrane on their north side, may be attributed to the Byzantine period and ourred probably before the fort s extension. Seondly, on the north-western and part of the south-western side of the struture, three towers and the rampart were roughly rebuilt with a simple and thin two-faed wall following the original plan. This rebuilding phase is very diffiult to plae hronologially. It is ertainly later than the other modifiations desribed above, but may still be Byzantine, or even early Islami or medieval. In spite of the bad quality of the masonry, it is more likely to date to a later period however. The addition of another, slightly smaller, U-shaped tower on the same side as this rebuilding may well be ontemporary to this phase. Two smaller doorways, on both sides of the gateway, are other modifiations to the original plan and are also very late, as shown by their thresholds, situated almost at the present-day ground level. From an arhitetural point of view, the original layout of the fort has good parallels in other late Roman fortresses. This is partiularly the ase with the muh larger legionary fortress of al-lajjun in Jordan, whih is dated to. AD 300 and shows similar fan-shaped orner towers and intermediary U-shaped towers (Parker 1991). In some ways, al- Bakhra /Avatha seems to be a small-sale version of al-lajjun. The fan-shaped orner tower is a feature ommon to most of the small forts built along the Strata Dioletiana. Although most of these forts have been asribed to the later Empire, only a ouple of them are preisely dated. Qusayr al- Sayla/Tetrapyrgium, for example, is demonstrated to have been built under Constantine I around AD 320 (Konrad 2001a, 153). It is important to bear in mind that it is diffiult to date a Roman fort by its plae in an hypothetial

9 DENIS GENEQUAND Al-Bakhra (Avatha), from the Tetrarhi Fort to the Umayyad Castle 233 Figure 9. Al-Bakhra : enlosure wall and towers of the Umayyad extension and north-east Tetrarhi rampart in the bakground (photo: D. Genequand). Figure 10. Al-Bakhra : irular orner tower of the Umayyad extension (photo: D. Genequand).

10 234 LEVANT typology alone. However, amongst the few typologial elements whih an be learly hronologially plaed, the development of the U-shaped tower appears to have ourred mainly in the eastern provines during the reign of Dioletian (Gregory , 166). Amongst several other well-dated monuments, the fortress at al-lajjun may be ited in addition to the fortresses at Luxor (El-Saghir et al. 1986), Nag al-hagar (Wareth and Zignani 1992) and Babylon/Old Cairo (Grossmann et al. 1994), three of whih are loated in Egypt. Despite the preliminary proposal of the exavator (Killik 1983), the still unpublished fortress of Udhruh, whih has a plan similar to al-lajjun and therefore very similar to al-bakhra, is most likely also to have been a Tetrarhi foundation. To summarize, the ombination of the epigraphi evidene (an insription whih is now loated in a seondary position), the stratigraphi evidene, and the arhitetural parallels, leads to the onlusion that the struture loated in the entre of the site of al-bakhra was originally built between AD 293 and 305 as the fort of an auxiliary avalry unit, the Equites promoti indigenae. extension and two possibilities may therefore be onsidered: either it was on the south-eastern side of the extension, whih is badly preserved in its entre due to erosion, or there was an entrane somewhere through the Roman rampart whih is ommon to both parts. If the latter was the ase, whih is more likely, a fairly low door overed by the sedimentary deposits may be supposed. It is of ourse to this addition to the original enlosure that someone investigating Umayyad astles would be drawn. Indeed, despite other examples of Roman forts where an extension was added or whih were somehow subdivided (Mons Claudianus: Maxfield and Peaok 1997; 2001; Abu Sayfi/Sile: Bonnet and Valbelle 2000, ), several elements suggest that the extension of al-bakhra was of early Islami date. These elements inlude firstly the semi-irular form of the towers, a shape unknown in Bilad al-sham before the Umayyad 5. The Umayyad extension A huge extension was built against the north-eastern side of the Roman fort (Fig. 2). It forms a seond retangular enlosure whih abuts the northern and eastern orner towers of the fort, and measures m. (exterior dimensions) (5517 m 2 ). The enlosure wall is thinner (1.80 m.) than that of the Roman fort. It is preserved up to 2.30 m. high and is built with small regularly ut bloks. It is reinfored by semi-irular towers, 8.65 to 9.85 m. in diameter, whih ontain retangular rooms (Figs 9 and 10). There is one exeption, the northern orner tower, whih has a retangular plan and does not seem to have been preeded by a semi-irular one. The entire area delineated by this new enlosure was built-up with rooms, or groups of rooms, organised along one or more north/south axes. Bearing in mind the kind of sedimentation, as well as the lak of any signifiant amount of stone debris, it is likely that most of the internal walls were built in mudbrik. Only the vertial monolithi posts of the doors, the lintels, and the arhes used for portioes and supporting roofs were built in stone. Numbers of door posts are therefore still standing, as well as some of the arhes whih are intat. There was only one group of rooms, loated in front of the southern intermediary tower, that seems to have been built entirely in stone. There is no visible entrane to the Figure 11. Al-Bakhra : sounding against the enlosure wall of the Umayyad extension, with remains of the mortar floor over a pebble bed (photo: D. Genequand).

11 DENIS GENEQUAND Al-Bakhra (Avatha), from the Tetrarhi Fort to the Umayyad Castle 235 Mosque Churh BAKHRA Plan de l église et de la mosquée Fondation Suisse-Liehtenstein pour la Reherhe Arhéologique Relevé et dessin: Sophie Reynard, Christian de Reynier Murs maçonnés observés Murs maçonnés restitués Puits et anaux Courbes de niveau: équidistane = 0,5m 0 10m N Figure 12. Al-Bakhra : plan of the hurh and ontiguous mosque (drawing: S. Reynard and C. de Reynier).

12 236 LEVANT period. This type of tower is for example found on the neighbouring sites of Qasr al-hayr al-sharqi (Large Enlosure) and Qasr al-hayr al-gharbi, as well as on several other Umayyad astles in southern Syria and Jordan. Seondly, ompared to the Roman rampart the thikness of the wall seems insuffiient to fulfill any effetive defensive funtion. Thirdly, there is the text of al-tabari whih indiates learly that one may expet to find some sort of Umayyad onstrution near, or linked to, the Roman fort of al- Bakhra. A sounding done against the enlosure wall of the extension, halfway between the northern square tower and the first intermediary tower, provided some evidene to support an early Islami date. The enlosure wall is preserved up to a height of 2.30 m. and due to the absene of any substantial amount of ollapsed stones, it is very likely that its upper part (not neessarily very high) was built in mud-brik. There is learly only one phase of onstrution and a single floor abutting the enlosure wall (Fig. 11). The floor is built in lime mortar over a pebble bed. This pebble bed only overs some thin layers of oupational debris overlying the natural limestone bedrok. Over the floor, the infill of the building onsists of remains of molten mud-brik and windblown deposits. Very little material was found in the sounding, but it seems to indiate a short oupational sequene. A handful of sherds originating from the floor (pebble bed and mortar) and from the immediately underlying layer provide a terminus postquem of at least the sixth entury AD. Of note amongst these sherds were several handles of Late Roman Amphora 1 (fifth to seventh enturies AD) and several sherds of possible early Islami date. The regional erami prodution during the Byzantine and early Islami period is still very poorly understood, but some of these sherds have parallels in the material from the well-dated Umayyad phase at Qusayr al-sayla/tetrapyrgium (Konrad 2001b) and from the Umayyad layers reently exavated at Qasr al-hayr al-sharqi (Genequand forthoming). 6. Churh and mosque Immediately to the east of the southern tower of the Roman fort are the remains of a hurh, against whih a later mosque was built (Fig. 12). The hurh is a relatively small building ( m. externally) whih follows the same orientation as the fort, roughly north-west to south-east. The identifiation of this struture as a hurh is based on its basilial plan with a entral nave and two side-aisles. A triumphal arh gives aess from the nave to a presbyterium, whih is flanked by two rooms (pastophoria) at the ends of the side-aisles. Roughly moulded imposts, whih supported the triumphal arh, an still be seen. There are several fallen olumns and a badly preserved orinthian apital in the entral nave. No apse is visible in the presbyterium. The hurh has two doors in the western and northern walls. It has also an annexe to the south. This plan, basilial with a tripartite santuary, ould only be of a hurh in suh a Late Antique ontext. The absene of a visible apse is not problemati. In the Syrian provines, a retilinear hevet is far more ommon. The entral part of the tripartite santuary (the presbyterium) very often features an insribed apse, but retangular plans are also attested, as at Khirbat Hassan in northern Syria, along with twenty-four other examples in the same area, or at Firja in the north-east (Sodini 1989, ). The retangular plan is therefore more ommon in northern Syria. It ould be that the masonry of an insribed apse at the hurh of al- Bakhra is not preserved up to the level of the urtain wall and present ground level. In relation to the hurh, it is worth mentioning that three ruiform monolithi basins were found sattered aross the site, whih may well be small baptismal fonts (Fig. 13). As there is no reason for suh a high number of them in al-bakhra itself, it may be that they are re-used elements brought from Palmyra, or elsewhere, during the early Islami period. Diretly to the north of the hurh are the remains of a mosque. This struture onsists of two rows of olumns and pillars defining an hypostyle hall ( m. internally) with five aisles and two, or possibly three, bays (Fig. 14). The entral aisle or nave is slightly wider than the side aisles. There is no disernible ourtyard. The qibla wall orresponds to the northern wall of the hurh. This wall is well preserved and there is no obvious mihrab in it. One possibility is that it was plaed in the northern door of the hurh, whih is almost in line with the entral aisle and whih appears to have been bloked by some later masonry. The other possibility is that it was a rather low nihe plaed in the middle of the entral aisle that has subsequently been overed by sedimentary deposits whih reah a height of more than 1.50 m. in this area. Re-using part of a previous building means that the qibla wall is not exatly oriented and faes slightly too muh to the south-west instead of a more orret south orientation. All the pillars and olumns supporting the roof of the prayer hall are re-used elements dating bak to the Roman period. The olumns inlude two milestones brought from the viinity (Bauzou 1993, 47). Another olumn bears a frame in relief, but without

13 DENIS GENEQUAND Al-Bakhra (Avatha), from the Tetrarhi Fort to the Umayyad Castle 237 Figure 13. Al-Bakhra : ruiform monolithi basin, probable baptismal font (photo: D. Genequand). Figure 14. Al-Bakhra : olumns of the mosque s prayer hall and hurh in the bakground (photo: D. Genequand). an insription. The eastern pillar of the southern row is a stele bearing the Tetrarhi Latin military insription already mentioned. At least two of the olumns of the southern row supported dori apitals, probably dating to the first entury BC or the first entury AD. Due to the relative rarity of the dori order in the Syrian provines, it would be surprising if they originated from al-bakhra itself and it may be more likely that they were brought from Palmyra like the other re-used arhitetural elements. Here again, the absene of a visible mihrab in this building whih has not been exavated is not a real obstale to its interpretation as a mosque. As pointed out, exavation along the qibla wall may bring it to light from under the sedimentary deposits in one of the two proposed loations. Rather than the presene or absene of a mihrab, however, it is the very regular plan defining an hypostyle hall whih indiates the struture s funtion, for whih the reused insription provides a general terminus postquem. The two rows of arhes departing from two pillars on the west wall and running parallel to the presumed qibla wall, form a harateristi element of mosque arhiteture, whih was presumably not used for just any kind of building. The date of the mosque is diffiult to establish with ertainty. Of ourse, one would expet to find an Umayyad mosque in al-bakhra, as is the ase for most known Umayyad astles. Its position diretly to the north of the hurh learly realls the similar situation of the ongregational mosque in Rusafa built under Hisham b. Abd al-malik (Sak 1996). However, its plan differs slightly from most of the other mosques assoiated with Umayyad astles in the sense that it is muh larger, with five aisles instead of the usual three aisles found, for example, at Jabal Says (Brish 1965, ), Hallabat (Bisheh 1980, 73 74), Umm al-walid and Khan al- Zabib (Bujard and Genequand 2001, ). Admittedly, all these monuments are situated in the southern part of Bilad al-sham (southern Syria and Jordan) and the mosque in al-bakhra may therefore perhaps indiate a different tradition in the north. However, it must be born in mind that a differene in size may also be related to a differene in site status. Qasr al-hayr al-sharqi and Rusafa (Grabar et al. 1978; Sak 1996), whih learly have a different and higher status than Umayyad astles, have large mosques with a ourtyard and seven and nine aisles respetively. This reflets the higher status of the settlements and is not a regional tradition. The unusual plan and large size of the mosque at al- Bakhra may therefore indiate that the site was more than simply a astle/qasr. We shall ome bak to that point below. It annot be ompletely ruled out that the mosque is of a later date than the Umayyad period. However, given the apparent later derease in importane of the site and onsequent lak of need for suh a monument, a later date for the mosque seems unlikely. Of relevane for both the hurh and the mosque is their position beside the fort s orners and lose to its gateway. The onstrution of the hurh beside one of the fort s orners is learly reminisent of a very similar situation at the fortress of Udhruh (Killik 1983, 113, 115), or, in a different ontext, at the sixth entury palae of Qasr Ibn Wardan (Maffei 1995). The palatial residene in Dumayr, of whih the date and history are still unertain

14 238 LEVANT (Byzantine or early Islami), offers another parallel of a military-like struture with an aompanying hurh built beside one of its orners (Lenoir 1999, 230, 233). In Umayyad omplexes, it is very ommon to find the mosque diagonally in front of the astle gateway. This is seen at Hallabat and Humayma (Oleson et al. 1999, ), where the mosques are situated exatly in the same way to the south-east of the gateway, or at Umm al-walid, Khan al-zabib and Qusayr Amra (Genequand 2002), where, despite a slightly longer distane between the astle and the mosque, the situation is similar. It is not possible, without exavation and absolute hronologial data for both strutures, to assess the relationship between the hurh and the mosque. The hurh ould have been abandoned before the onstrution of the mosque. At the same time, as the onstrution of the latter does not enroah at all on the former, exept by possibly bloking its northern door, they ould have funtioned simultaneously for a while. Historially or religiously, both solutions are aeptable and further fieldwork would be neessary to provide a more aurate idea of that partiular point of the religious life and development of the site. 7. Houses More than twenty houses are distributed aross the south-western, southern and eastern parts of the site (Fig. 1). Some of them have plans that are still easily disernible on the surfae. They usually onsist of retangular strutures omprising a ourtyard and a series of rooms arranged on one, two or three sides of the ourtyard. A onspiuous arhitetural element of these strutures is the frequent use of monolithi door posts, also seen inside the fort and its extension. A ouple of stone doors were also observed lying elsewhere on the site, one of whih originated from an early Empire Palmyrenian tomb. Without exavation, it is not possible to date preisely the onstrution and oupation of these houses or to know if they relate more to the Late Antique or early Islami settlement. Al-Tabari learly mentions a village (qariya) beside the astle at the time of al-walid b. Yazid s murder and it may therefore be assumed that at least some of the houses were being used during the first half of the eighth entury. All six of the wells mentioned above were situated in the two areas overed by houses, two in the east, three in the south and one between the fort s gateway and the hurh. Similarly, the probable baptismal fonts re-used as basins were found in these two areas. Two aqueduts distributing water from the artesian spring in a southern diretion and along the south-eastern side of the fort, through the help of a now destroyed water-lifting installation, are of a muh later date, in other words, medieval or more probably modern. 8. Funerary strutures There are numerous funerary strutures in and around the site. The best preserved are small monuments ontaining several individual tombs. One, probably of Byzantine date, is in front of the fort s gateway, about 50 m. to the east. It has a entral retangular funerary hamber surrounded by five tombs built as sarophagi into the walls. In the eastern part of the settlement, there are two square monuments of whih the bases ontain several burial ompartments. Both of these monuments are related to an enlosure defined by a low wall. The southern part of the site ontains a number of groups of retangular tombs and in a roky outrop, outside the settlement and further away to the south, three parallel rok-ut tombs are visible. There are also two large neropolis loated further to the east and to the south-west of the site. Both of them have been ompletely plundered, the eastern one between 1908 and 1912 (Musil 1928, 140, 286). The tombs of the southern neropolis are very damaged and even their orientation is diffiult to assess. They onsist mainly of large slabs enirling holes whih are now filled in by sand. In the eastern neropolis, the situation is slightly better. Despite their robbing, the tombs appear to have been oriented east/west and most of them had small vertial stelae at both extremities. This more losely reflets Muslim praties, but Byzantine or Christian use of this neropolis is also shown by two stelae with engraved rosses. Both of these emeteries may therefore have been related to the Late Antique and early Islami oupation of al-bakhra. A loser examination of these funerary strutures was neessary beause of the statement by al- Masudi (Muruj, 4, 49) that al-walid b. Yazid, after having been killed in al-bakhra, was buried there. Altough some of the tombs were Muslim, espeially in the eastern neropolis, no partiular tomb or monument ould be related to that event. Indeed, for suh an early period in Islami funerary arhiteture, one would not expet to find an ostentatious monument, the more so beause al-walid b. Yazid was never partiulary venerated after his death. Furthermore, it is known that the Abbassids destroyed the tombs of the former Umayyad aliphs

15 after they took power. The best examples of Umayyad tombs whih ould serve as omparisons to what we would expet to find in al-bakhra, are those of the neropolis in al-qastal in Jordan (Imbert 1992). These tombs were retangular pits at the extremities of whih were plaed insribed stelae. 8. Enlosures The entire northern part of the site is oupied by a network of low walls forming about fifty enlosures (Fig. 1). They are mainly of retangular form and follow roughly the same alignment as the fort. The enlosures are lose to eah other and shared walls are ommon. A few paths separate them and give aess to the enlosures whih are in the entre of the network. The bedrok is visible inside several of them and where it is not, it does not seem to be buried under a signifiant amount of sediment. It therefore appears that these enlosures were not intended for ultivation but were probably for penning attle. The enlosures are diffiult to date and may be related to any period of oupation of the site. They orrespond well with the use of the Roman fort by a avalry unit, whih would have to pen numerous horses, or with a later and more agriultural settlement partly direted towards the breeding of attle or horses. The absene of any mention of similar features near a Roman fort in the Near East on suh a large sale makes the seond possibility more likely. 10. Disussion Al-Bakhra is one of the very few Umayyad sites for whih the historial soures provide more details than the name of the plae and the owner s name. Therefore, despite the lak of any large-sale arhaeologial exavation, it is worth attempting a better understanding of the settlement by bringing together the textual and arhaeologial evidene. Al-Tabari s aount makes the differene very lear between three elements: a village (qariya); a tent or amp (fustat); and a astle, equally alled qasr or hisn, where, as a last resort, the aliph al-walid b. Yazid took refuge before being killed. It is not obvious whih one of the two parts revealed by the arhitetural study orresponds to the qasr of Nu man b. Bashir. It ould be that it was the extension. If this was the ase, the anient military fort ould have been some sort of village in a fortified enlosure. This development mirrors the fate of other Roman forts in Arabia, suh as Umm al- DENIS GENEQUAND Al-Bakhra (Avatha), from the Tetrarhi Fort to the Umayyad Castle 239 Rasas/Kastron Mefaa (Bujard 1995; Piirillo and Alliata 1994) and Udhruh/Augustopolis (Killik 1983). However, it is equally possible that the qasr was loated inside the Roman enlosure and that the re-used apitals seen by Musil, whih ertainly originated from Palmyra, were employed in one or several ourtyards with portioes or hypostyle halls of Umayyad date. There are learly very few Roman forts re-used as Umayyad astles. Even though some surveys have laimed a re-oupation of a Roman fort by the Umayyads on the basis of surfae sherds, there is atually no real evidene for most, if not all, of these instanes. The only exeption is Hallabat in Jordan, where the extent of the reonstrution is not absolutely lear from the available material (Bisheh 1980; 1982), but seems to have involved major modifiations to the original struture. As to be expeted, and as is amply demonstrated by later Islami praties in palatial building ativities (Samarra for example), eah prine or ruler built his own palae(s), a fat that explains the rarity of Roman forts being re-used by the Umayyads and their entourages. This last point may indiate that the qasr was formed only by the extension, whih is an ex nihilo onstrution and whose surfae area (5517 m 2 ) orresponds roughly to that of a lassial square Umayyad astle with eah side measuring about 70 m. (4900 m 2 ). On the other hand, if we onsider that the qasr omprised both the Roman fort and the extension, in other words a new ex nihilo onstrution built against an older one with a probably muh modified internal layout, it would imply that it was in fat related to a different ategory of buildings. Its surfae area (20,489 m 2 ) would have been far greater than any other Umayyad astle and omparable to the new urban settlements like the Amman itadel (15,876 m 2, only the palae area), Aqaba (23,100 m 2 ), the Large Enlosure in Qasr al-hayr al-sharqi (27,889 m 2 ) and even the muh larger Anjar (114,700 m 2 ) (Northedge 1994, ). However, this possibility is unlikely beause all of these urban settlements were entirely new onstrutions, ontrasting strongly with the re-used Roman fort. In addition, the sounding against the Tetrarhi rampart does not provide any deisive evidene of important Umayyad building ativity. However, it should be kept in mind that the sounding was small and that there are other areas inside the fort that provide arhitetural indiations of modifiations or rebuilding ativities that orrespond well with the Umayyad tradition (for example portioes with reused orinthian apitals). Moreover, the loation of

16 240 LEVANT the mosque, like that of the older hurh, near the gateway of the Roman fort shows that the latter remained the main entrane during the Umayyad period. There is no reason to aept the interpretation of Musil, who explains the qasr as an integral part of the fustat, as well as being the strongest one (Musil 1928, 295; hene Sourdel-Thomine 1960). Al- Tabari s text remains vague but gives enough lues to interpret the qasr and the fustat as two ompletely independent elements. The fustat, in the view of the present author, is a ampsite of tents where the aliph, his entourage, and probably his soldiers, stayed after they arrived in al-bakhra (al-tabari, Ta rikh, II, 1803). This view is supported by the desription of the plundering of Walid s amp after his death, in whih the amp is designated as askar, literally meaning the army, i.e. the amp of the army (al-tabari, Ta rikh, II, 1807). In this ontext, the word does not give the impression of a permanent onstrution and, despite the presene of numerous enlosures potentially used as animal pens, should not be interpreted as being one of the garrison settlements ( askar, mu askar, ma askar) used as stables and military stud-farms in the aliphal period (for this latter definition see Viré 1978, 1144). The most satisfatory solution would thus be to see al-bakhra as something in between a single qasr and a new urban settlement, a solution whih is perhaps indiated by the textual soures and whih has already been suggested in relation to the mosque. Leaving aside the problem of the fustat, whih is definitely something temporary and separate from the solidly built omponents of the site, there remains the problem of differentiating between a astle/qasr and a village/qariya. It is lear that the houses sattered around the Roman fort and the extension are, if they date to the Umayyad period, part of the village. With no onlusive evidene that loates the qasr between the two enlosures, it would seem more likely that both enlosures formed the qasr, bearing in mind that only part of its surfae area would have been the palatial area or the main owner s residential area. The latter would only have been loated in one of the two enlosures, or would have partially oupied both, the rest being servie buildings and owner s relatives or entourage houses. This possibility therefore gives the Umayyad settlement a proto-urban aspet whih heralds the more formally planned new urban settlements suh as those at Qasr al-hayr al-sharqi, the Amman itadel, and Anjar. If the site really belonged to Nu man b. Bashir, and there is no reason to doubt this attribution, it is to be expeted that it would differ from the other Umayyad astles or settlements whih are well known and whih are all dated to the first half of the eighth entury AD. Nu man b. Bashir was one of the Companions of the Prophet (Zetterstéen 1995). He was governor of Kufa and, at the end of his life, of Hims. He was killed shortly after the battle of Marj Rahit in AH 65/AD 684, a date whih provides a terminus antequem for the Umayyad onstrutions at al-bakhra, or at least for some of them. Al-Bakhra thus represents one of the oldest Umayyad astles or related settlements, from a time where a more standard plan (if there is really one) had not yet been definitely established. 11. Conlusion An arhaeologial and arhitetural survey followed by a short season of soundings in al-bakhra have brought to light a somewhat unexpeted and unique example of a Late Antique military fort transformed into an Umayyad settlement of a type very different to those that are already well known. It is hoped that future work in al-bakhra will improve and add to our understanding of the site and of its omponents. Two important observations about the site were aquired in the present projet. First, the Roman fort, whih is the first one to be exavated along the Strata Dioletiana and other roads between Damasus and Busra to the south and Rusafa to the north (Palmyra being the exeption) is definitely a onstrution of Dioletiani date. Its plan, a redution of the larger ontemporary legionary fortresses, will serve as an important benhmark for a renewed study of other forts and amps along the roads of the Syrian desert. Seond, the extension of the fort an be safely onsidered as an Umayyad addition. Together with other arhaeologial evidene (e.g. the presene of a mosque and early Islami eramis), this gives a definitive indiation of the site s identifiation with the al-bakhra of the written soures relating the murder of the aliph al-walid b. Yazid. The hypothesis presented onerning the form and type of the Umayyad settlement, whether orret or not, thus provides a solid base for ontinuing the exploration of al-bakhra. Aknowledgements The Umayyad Settlement Projet is a joint Syrian-Swiss projet between the Diretorate General of Antiquities and Museums (DGAM) and the Swiss-Liehtenstein Foundation for Arhaeologial Researh Abroad (SLFA - Zürih). The projet is o-direted by Denis Genequand

17 DENIS GENEQUAND Al-Bakhra (Avatha), from the Tetrarhi Fort to the Umayyad Castle 241 (IASA/Lausanne and CBRL/Amman) and Walid al-as ad (DGAM/Palmyra). We would like to thank the Swiss-Liehtenstein Foundation for Arhaeologial Researh Abroad and the members of its board and sientifi ommittee for supporting and funding the projet. We would also like to thank the Diretorate General of Antiquities and Museums of Syria for weloming us and for help and assistane, espeially the former General Diretor and now Deputy Minister of Culture Dr Abd al-razzak Moaz, the General Diretor Dr Tammam Fakouh and the Diretor for Exavations and Arhaeologial Studies Dr Mihel al-maqdissi. The team whih worked at al-bakhra inluded during both seasons: Cyril Ahard (student in arhaeology, University of Paris IV), Hugo Amoroso (student in arhaeology, IASA/Lausanne), Sylvain Dumont (surveyor, Paris), Maria Haldemann (student in arhaeology, IASA/Lausanne), Jamil Hassan (arhaeologist, DGAM/Palmyra), Sophie Reynard (surveyor, Paris) and Christian de Reynier (arhaeologist, SPMS/Neuhâtel). In Palmyra, we have benefitted from the help and assistane of Khaled al-as ad and Walid al-as ad, respetively former and present Diretors of the Palmyra Offie of the DGAM. For helping with the organisation and for material, we are grateful to the CBRL in Amman, to whih the projet is affiliated, and to the IFAPO in Damasus. Dr Marie-Odile Rousset (Lyon) kindly disussed with us the pottery from the exavation. Speial thanks are due to Professor Charles Bonnet, Professor Alastair Northedge and Professor Daniel Paunier for enouragement, advie and support. Grants from the Soiété Aadémique Vaudoise and the van Walsem Foundation, both based in Lausanne, allowed us to ontinue researh on the Umayyad Near East in the best onditions at the CBRL in Amman. Finally, we must thank Samantha Dennis, Dr Charlotte Whiting and Dr Bill Finlayson for helping with the English text. Bibliography Primary soures Al-Bakri, Abu Ubayd Abd Allah ibn Abd al- Aziz, Mu j. am ma ista j. am. Ed. F. Wüstenfeld. Deuerlih she Buhhandlung and Maisonneuve & Comp.: Göttingen and Paris, Ibn Qutayba, Abu Muhammad Abd Allah b. Muslim al- Dinawari, Kitab al-ma arif. Ed. T. Akasha. Dar al-ma arif: Cairo, Al-Masudi, Abu l-hasan Ali b. al-husayn, Kitab Muruj al-dhahab wa-ma adin al-jawhar. Ed. C. Pellat. Publiations de l Université libanaise: Beirut, 5 vols, Al-Masudi, Abu l-hasan Ali b. al-husayn, Kitab al- Tanbih wa l-ishraf. Ed. M.J. De Goeje, BGA, VIII. Brill: Leiden, Notitia Dignitatum. Ed. O. Seek. Weidmann: Berlin, Al-Tabari, Abu Ja far Muhammad ibn Jarir, Ta rikh alrusul wa-l-muluk. Ed. M.J. De Goeje et al. Brill: Leiden, Yaqut, ibn Abd Allah al-hamawi, Mu j. am al-buldan. Ed. F. Wüstenfeld. Brokhaus: Leipzig, Seondary soures Bauzou, T. (1989) A finibus Syriae. Reherhes sur les routes des frontières orientales de l Empire Romain. Unpublished PhD thesis, Université de Paris I. (1993) Epigraphie et toponymie: le as de la Palmyrène du Sud-Ouest. Syria 70, Bisheh, G. (1980) Exavations at Qasr al-hallabat, ADAJ 24, (1982) The Seond Season of Exavations at Hallabat, ADAJ 26, Bonnet, C. and Valbelle, D. (2000) Le amp romain de Tell El-Herr dans l arhiteture militaire du Bas- Empire. Pp in D. Valbelle and J.-Y. Carrez-Maratray (eds) Le amp romain du Bas-Empire à Tell El-Herr. Errane: Paris. Bujard, J. (1995) La fortifiation de Kastron Mefa a/umm ar-rasas. SHAJ 5, Bujard, J. and Genequand, D. (2001) Umm al-walid et Khan az-zabib, deux établissements omeyyades en limite du désert jordanien. Pp in B. Geyer (dir.) Conquête de la steppe et appropriation des terres sur les marges arides du Croissant fertile. Maison de l Orient Méditerranéen: Lyon (TMO 36). Brish, K. (1965) Das omayyadishe Shloss in Usais. II. Mitteilungen des Deutshen Arhäologishen Instituts, Abteilung Kairo 20, Genequand, D. (2002) Une mosquée à Qusayr Amra. ADAJ 46, (2003) Projet «Implantations umayyades de Syrie et de Jordanie». Rapport de la ampagne de prospetion (juin-juillet 2002). Shweizerish-Liehtensteinishe Stiftung für Arhäologishe Forshungen im Ausland (SLSA/FSLA/SLFA) Jahresberiht 2002, Zürih, (forthoming) Rapport préliminaire de la ampagne de fouille 2003 à Qasr al-hayr al-sharqi (Syrie). Shweizerish-Liehtensteinishe Stiftung für Arhäologishe Forshungen im Ausland (SLSA/FSLA/SLFA) Jahresberiht 2003, Zürih. Grabar, O., Holod, R., Knustad, J. and Trousdale, W. (1978) City in the Desert: Qasr al-hayr East. 2 vols. Harvard University, Center for Middle Eastern Studies: Cambridge, Mass. Gregory, S. (1995 7) Roman Military Arhiteture on the Eastern Frontier. 3 vols. Adolf Hakkert: Amsterdam. Grossmann, P., Le Quesne, Ch. and Sheehan, P. (1994) Zür römishen Festung von Babylon Alt-Kairo. Arh Anz 1994, Imbert, F. (1992) La néropole islamique de Qastal al- Balqa en Jordanie. Arhéologie islamique 3, Kennedy, H. (2001) The Armies of the Caliphs. Routledge: London. Killik, A.C. (1983) Udruh. The Frontier of an Empire : 1980 and 1981 Seasons, a Preliminary Report. Levant 15,