Hammat al-Qa and the roots of urbanism in southwest Arabia.

EDENS, C. ; WILKINSON, T.J. ; BARRATT, G. 等

Arabia is not known for its early towns and there are few records of urban-scale settlements before the 1st millennium BC. Our aim here is to announce the existence of numerous large settlements of the 3rd and 2nd millennia BC in highland Yemen, and to illustrate the layout of one of the best examples. In 1999 the site of Hammat al-Qa was planned stone-by-stone to reveal a complex and dense pattern of buildings over a total site area of 5 ha, together with off-site features such as terraced fields and threshing floors. Chronological control is provided by radiocarbon assay of excavated charcoals and a local ceramic sequence. The site will be placed in the broader context of urbanism in Arabia in order to highlight the significance of the site, and to show how it differs from comparable large settlements found elsewhere in the peninsula.

By the end of the 3rd millennium BC large settlements were evident at Tarut Island in Saudi Arabia (Potts 1990: 66-7, 105-9), Qalat al-Bahrain (Hojlund & Andersen 1997), Saar (also on Bahrain: Crawford et al. 1997), Tell Abraq (UAE: Potts 1991), Hili in the UAE (Cleuziou 1982), and Bat in Oman (Frifelt 1976) (FIGURE 1). Qalat Bahrain, a tell measuring 700 x 400 m, is the largest of these settlements, and if occupied to its fullest extent in the late 3rd millennium, this must have been a town capable of housing some 3000 people (Hojlund 1989). Unfortunately owing to the overburden of later occupation, the layout of this as well as other large sites on Bahrain is little known, and we must therefore turn to sites such as Saar for evidence of settlement layout (Crawford et al. 1997). Similarly, in interior Oman and the Emirates although settlements may have been extensive, their location within cultivated oases has resulted in their plans being obscured or disturbed by later agricultural activities (Cleuziou 1982). Thus Frifelt's (1976) estimation of 30-40 ha for Bat is based on an assumed continuity of settlement between fortified towers; if only extant building remains are used the settlement decreases to around 4 ha in area (Brunswig 1989).

[Figure 1 ILLUSTRATION OMITTED]

In Yemen pre-Iron Age settlement was first documented in 1981 in the Khawlan district of the eastern highlands, where an Italian team investigated numerous small 3rd-millennium settlements (de Maigret 1990). Similar small sites sharing the same architectural and ceramic tradition appear in the desert fringes of interior Yemen (e.g. Blakely et al. 1996; Cleuziou et al. 1992; Breton & Darles 1994). Finally, at Sabir near Aden recent investigations have demonstrated the existence of a massive town of the late 2nd millennium and early 1st millennium BC (Vogt & Sedov 1998).

In the Yemen highlands, where rainfall in excess of 300 mm per annum fails on intermontane plateaus between 2000 and 3000 m above sea level, recent archaeological surveys indicate that large Bronze Age settlements were common. Unlike the settlements of the Gulf and southeast Arabia, those of Yemen are on hilltops and are distinguished by the clarity of their building layouts which consist of collapsed and heavily weathered walls. A key factor for settlement is that as a result of their location in the zone of summer monsoonal rainfall these highlands are the best watered part of the Arabian peninsula. The potential for significant crop production is high as long as there is sufficient soil for cultivation, a deficit that is usually addressed by the construction of terraced fields.

The Dhamar survey

Since 1994, a team from the Oriental Institute, University of Chicago has surveyed and conducted limited excavations in the intermontane plains and surrounding highlands of the Dhamar region, some 80 km south of Sana'a in Yemen. As this part of Yemen had not previously received systematic study, the project has developed a ceramic sequence using comparisons with pottery from the drier eastern highlands (Khawlan, Jawf, Wadi al-Jubah, Wadi Markha) and radiometrically dated excavated samples (Edens 1999). To date the survey has recorded 56 Bronze Age sites, several of which exceeded 3 ha in area. Noteworthy sites include: Kharraib (DS 228: c. 3 ha), Medinet al-Balad (DS 187: c. 4 ha), Sibal (DS 66: c. 2 ha), Hawagir (DS 293: 10-15 ha), al-Mosayn'a (DS 268: c. 2.5 ha), and Hammat al-Qa (DS 101:3-5 ha) (FIGURE 1).

Around Dhamar sites occur mainly in the semi-arid highlands where today mean annual rainfall is between 250 and 350 mm. Rainfall may however have been slightly higher during the 3rd millennium BC when monsoonal activity, although declining, was stronger than today. In the northeast part of the survey area where rainfall is less, scattered Bronze Age villages are evident, whereas to the southwest where rainfall is higher Bronze Age sites are less visible. This is, in part, because in the wetter highlands where field terraces extend to hill summits, sites have been transformed as a result of processes such as the removal of stones for terraced walls of fields.

The Bronze Age sites are usually shallow and lack later occupation. Where adverse taphonomic effects have not occurred, the stone architecture of many sites is readily apparent and therefore large portions of these sites can be mapped. Hammat al-Qa provides an exceptionally well-preserved example of this kind.

Hammat al-Qa

Hammat al-Qa was discovered by analysis of aerial photographs prior to the 1995 field season. The survey team visited Hammat al-Qa in 1995, and dated the site to the 3rd millennium BC by the comparison of surface pottery with that from the Khawlan sites (Wilkinson et al. 1997). The site sits upon a gently inclined but otherwise flat-topped hill the summit of which is 50 m above the level of a narrow plain. This surface is the remnant of a basalt flow that overlies, in turn, a conglomerate of alluvial cobbles and a basic igneous formation. The basalt cap provided building material for the town, and the conglomerate felsite cobbles for the chipped stone industry.

During the 1996 season preliminary mapping identified the basic architectural components of the site, and five soundings acquired pottery and radiocarbon samples in association with the town wall and several buildings. Two additional soundings and detailed mapping of the entire site and its surrounding area in 1999 completed the field investigation to date. These investigations date the site to the range 2320 to 1410 cal BC (TABLE 1), i.e. a little later than the Khawlan sites (de Maigret 1990) and Sibal (near Dhamar), and somewhat earlier than Kharraib and Hawagir, also in the Dhamar area (Wilkinson & Edens 1999; Edens 1999). A sounding against the southwest town wall revealed two phases of construction, separated by a spread of rubble. The rubble contained a mix of Bronze Age and later pottery, suggesting that the wall was rebuilt early in the 1st millennium AD or even later. Occasional Iron Age and Himyarite sherds on the surface, a single 1st millennium Bc radiocarbon determination, and the apparent remodelling of some architecture, all indicate late re-use of the site. The excavations provided a faunal assemblage of sheep/ goat (90%) and cattle (10%), but recovered no archaeobotanical remains besides rare wood charcoal. The Italian investigations in Khawlan report wheat, barley, oats, sorghum, millet and dates as Bronze Age crops (Costantini 1990).

TABLE 1. Radiocarbon determinations from the 1996 soundings at Hammat al-Qa.

 material
site and locus dated lab. no.

DS 101: Op 4/5 charcoal Beta-101071
DS 101: Op 3/6 charcoal Beta-101070
DS 101: Op 3/2 charcoal Beta-101069
DS 101: Op 4/8 charcoal Beta-101072

 date BC
 date BP calibrated
site and locus (uncalibrated) to 2 s.d.(*)

DS 101: Op 4/5 2590 [+ or -] 60 830-530
DS 101: Op 3/6 3260 [+ or -] 60 1670-1410
DS 101: Op 3/2 3510 [+ or -] 60 1960-1670
DS 101: Op 4/8 3730 [+ or -] 70 2320-1920

(*) calibrated with the CALIB 3.0.3 atmospheric model using the bidecadal tree ring data set.

Architectural fabric

Mapping in 1999 proceeded in two-stages. Initially, a total station was used to plot alignments of walls, loessic deposits, saghiyahs (water diversion channels), and agricultural terraces, as well as to fix topographic spot heights. This work created `strings' of points representing buildings or parts thereof, each point being marked in chalk at the appropriate spot of the wall (FIGURE 2). The second stage entailed plotting individual stones contained within each string, using a print-out of the string framework for orientation and tapes for measurement. Both stages of mapping often required judgment to pick out wall lines amidst the sometimes continuous spreads of wall collapse; the second stage, when only the component stones ot walls were plotted, acted as a useful corrective to the string framework.

[Figure 2 ILLUSTRATION OMITTED]

In addition to being the prime settlement of a two-tier hierarchy, the site exhibits several features reminiscent of urbanism, namely: a surrounding wall with gates, a dense layout of buildings, streets, and a possible subdivision into quarters (FIGURE 3). The defensive perimeter wall is visible almost continuously around the southeastern half of the hill but only intermittently around the northwestern perimeter. This wall is three irregular courses wide, 1.8 m thick and preserved in places 3-4 courses high; the stones are up to a metre long and unshaped. A cross-wall, 1.4 m thick with a rubble fill between a facing of large stones, runs northeast-southwest across the hill, cutting the site into two unequal parts (FIGURE 3). The dense architecture of the main site covers the larger southeast area of 3.0 ha, while a thinner scatter of buildings characterizes the remaining 1.9 ha of the hill (the northwest site). Ash middens that run up against the interior face of the perimeter wall on the northeast side of the main site provided two early-mid 2nd millennium BC radiocarbon dates (Operation 3: FIGURE 3: T3), firmly relating this wall with the Bronze Age settlement. Although the cross-wall, with its different construction style, has not yet been directly associated with the town, the adjacent Bronze Age structures were not robbed of their stones for construction of the wall, strongly suggesting that it too belongs to the Bronze Age settlement.

[Figure 3 ILLUSTRATION OMITTED]

Two gates offer access through the northeast perimeter wall. Both are framed by large orthostats, the more northerly gate 1 being 3.2 m, and the more southerly gate 2 1.5 m, wide (FIGURE 3:G1 and G2). A less obvious gate punctuates the southwest perimeter wall (G3), in which a stairway of several risers leads up to a gap in the wall. Since this part of the wall was rebuilt long after the Bronze Age, this gate does not certainly belong with the Bronze Age settlement.

The architecture within the main site is dense across its southeast two-thirds, thinning significantly to the northwest; the western corner is largely devoid of structures. Despite first impressions of disorder, built-up zones alternate with streets and open spaces to impose a perceptible order to the site as a whole. A wall curves westward from the perimeter wall to the cross-wall, closing off the northern corner of the town, the highest point of the butte (FIGURE 3: W). A street runs along the outside face of this wall from gate 1, and a topographic indentation half way along the wall may mark the entrance to this upper compound from the street. A strip of buildings on low terraces extend along the perimeter wall southeast of gate 1, forming a distinct zone about 80 m long and 25 m deep. In contrast, the southwest perimeter wall borders a large zone of open space in the northwest quadrant of the site, and a scatter of buildings farther south.

Between these two sides of the town, seven elongated blocks of housing 40-60 m by 25-30 m, alternate with streets and strips of open space, all being oriented with the site slope. Open strips, 10-15 m wide and devoid of architectural rubble, separate the northwestern blocks while northeast-southwest streets run between the remaining blocks; northwest-southeast streets delimit the blocks from architecture along the northeast and southwest perimeter wall. The streets generally have gravel and fine rubble surfaces, and boundaries of low curbing or retaining walls.

The combination of disaggregation and aeolian sedimentation obscures additional houses: therefore the assessment of 46-57 visible houses almost certainly under-estimates the original number (TABLE 2). These rectilinear structures are mainly oriented northeast-southwest with the site slope, and some are linked by retaining walls or lines of stones that run perpendicular to the slope. These linkages, together with the arrangement of buildings within blocks, hint at the existence of supraresidential social units. One block in the central main site offers a view of this organization. Here 6-7 houses are grouped around a central open space (FIGURE 4: b), with a lane entering the open space from the street to north (a), and three retaining walls (a stepped terrace?) connecting the two houses at the southwest end of the block (c).

[Figure 4 ILLUSTRATION OMITTED]

TABLE 2. Hammat al-Qa. Population estimates based on extant architecture.

 min. max. min. max.
 no. of no. of pop. pop.
 houses houses 5/house 5/house

Main town 46 57 230 285
northwest town 16 22 80 110
total site 62 79 310 395

 house house pop. est.
 area area 10 sq. m/ using
 est. 1 est. 2 person est. 2

Main town 2960 3534 296 353.4
northwest town 957 1144 95.7 114.4
total site 3917 4678 391.7 467.8

Outside the cross-wall to the northwest, buildings are sparse with relatively dense clusters occupying the western point of the hill, the eastern corner near the cross wall, and the central area. The layout of individual houses is far less obvious than in the main site, and there is less wall collapse and rubble. The absence of stones and rubble suggests that the component stones have been robbed out, perhaps for use in other parts of the site. The visible coherent architecture represents some 16-22 buildings.

Unambiguously complete plans are available for 16 houses in the main site and another 7 for the northwest site. Almost all the complete houses are simple rectangular or `L' -shaped structures, the latter appearing slightly more often than the former. The rectangular house comprises either a single room or is sub-divided into several rooms by cross-walls. Buildings are narrow (2-5 m), but may be; up to 30 m in length. Outer walls are of basalt blocks laid in courses or slabs set on edge; they are usually one course wide resting upon a foundation of smaller stones. Entrances are near the centre of the long southeast wall and usually are framed by orthostats. Small square or rectangular box-like annexes several metres across frequently are attached to the long northwest wall.

The houses vary considerably in size, and the complete structures range from 20 to 112 sq. m. The rectangular houses are larger than the `L'-shaped houses (averaging 64 and 52 sq. m respectively); houses in the main site tend to be larger than those of the northwest site (62 [+ or -] 28 versus 52 [+ or -] 26 sq. m) but this difference is not statistically significant. Several of the largest buildings (all rectangular) seem to include a walled open space at one end, a feature that also occurs in an excavated house at the late 2nd millennium site of Kharraib (Edens 1999).

Additional small structures associated with the houses include square or more frequently circular structures 1-3 m across, which sometimes abut the house wall. These structures resemble stone-lined hearths or possibly silos excavated at Bronze Age sites in the Khawlan and Dhamar areas (de Maigret 1990; Wilkinson & Edens 1999). However, not all of these structures belong with the Bronze Age settlement and some are demonstrably later.

Larger curvilinear walls roughly constructed of boulders often frame patches of aeolian silts in a terrace-like fashion. Soundings show that the silt of one such `terrace' covers Bronze Age walling and wall collapse and that similar silts bury occupation deposits within houses. Such boulder walling probably represents a post-Bronze Age remodelling of the site in an effort to retain the silts for agrarian uses. Several carefully constructed curvilinear walls, which extend downslope from groups of Bronze Age houses at the western end of the main site, may have been fenced enclosures for animals or for outdoor domestic activities attached to the Bronze Age residences.

The agricultural system

Since the first field season in 1994 it was apparent that the Bronze Age sites of the highlands were associated with relict terraced fields. Although such systems cannot be dated directly, the Hammat al-Qa relict fields form a coherent set of features that are tentatively suggested to have been in use during the Bronze Age occupation of the site. Such an early date for terraced agriculture in the area is not surprising given the recent discovery of a valley bottom terraced field buried below some 6 m of valley fill and dated by radiocarbon dated charcoal from the associated soil to 4960 [+ or -] 80 BP (Beta-117431 3955-3630 cal BC; Wilkinson & Edens 1999: 3).

The agricultural features at Hammat al-Qa are on stony hill slopes which appear originally to have been covered by a thicker humic loam palaeosol. Relict terrace walls appear as horizontal lines of heavily weathered stones, which are now eroded virtually flush with the slope. Excavation in 1996 showed these to have the characteristic features of field terraces, namely a terrace wall face and a supporting mass of fist-sized stones immediately upslope. Such relict terraces remain on the southwest, east and north-facing slopes of the hill (FIGURE 2). The two last named groups are situated near the base of the slope where they would have benefited from runoff received from the large area of hill slope above. In contrast terraces on the southwest-facing slope were half way up the slope, where they would have received additional runoff from the hilltop, thus rendering unnecessary a long catchment on the slope itself.

Additional off-site features include eroded masses of large, heavily weathered stones, 20-25 cm across, which appear to have formed flattened platforms on the northwest and southeast spurs (FIGURE 2: Th). By analogy with modern installations, and relict features associated with sites of the 1st millennium BC and AD, these platforms are interpreted as threshing floors. They are situated on eminences that would receive winds suitable for winnowing grain, and would have been conveniently located for the fields themselves (FIGURE 5).

[Figure 5 ILLUSTRATION OMITTED]

Although the relict field terraces are insufficient to allow for the reconstruction of the Bronze Age zone of cultivation, the number and area of buildings within the site can be used to estimate site population (TABLE 2), and in turn the area of cultivation needed to sustain that population. This population of roughly 300-500 persons would have required an area of land of only 1 km radius, thereby leaving the lowest part of the surrounding plain as potential pasture. Bronze Age land use around Hammat al-Qa, and in other basins, can therefore be suggested to comprise:

a hill slopes adjacent to the settlement where some cropping and crop processing took place;

b lower slopes and plain margins devoted to cultivated crops;

c bottom lands of basins which remained as pastoral reserves.

Discussion

The perimeter wall, structured fabric and architectural density indicate that Hammat alQa had some characteristics of an urban layout, despite its relatively small population. Although these characteristics are clearest at Hammat alQa, sites of a similar scale are relatively common in the central Yemeni highlands. When compared with present-day Yemeni settlement Hammat al-Qa can be seen to fall within one of the commonest size classes, namely the small rural town or large village with populations in the range 100-500, of which there were 11,000 in 1979, holding some 44% of the population (Steffen 1979: II/27). Unlike the 1st-millennium BC towns of the desert fringe such as Shabwa, or Late Bronze Age Sabir on the Indian Ocean plain, the buildings at Hammat al-Qa show very little variation in size or apparent status. On the other hand the settlements of Khawlan, just 50-60 km to the northeast, make a strong contrast in a different way (de Maigret 1990). These sites are smaller (none exceeding 1 ha) and contain compounds of contiguous curvilinear rooms set around a central space; this architectural form seems to reflect lineage residential groups. The architecture and the significantly drier environment of the eastern highlands suggest that the Khawlan settlements belonged to communities that placed a greater emphasis on animal herding and opportunistically farming. However, the evident contrasts with Hammat al-Qa may mask similarities of social organization, with the linkages between houses and their arrangement around a common space within blocks serving as the seats of extended kin groups within the settlement. These groups probably would have pooled labour in the creation, use and maintenance of agricultural terraces, threshing floors and other facilities.

In contrast with Bronze Age towns of the Gulf coast and southeast Arabia, where artefactual assemblages demonstrate numerous links with other parts of the Gulf, Iran, Mesopotamia and the Indus, the sites of the Yemen highlands show little evidence for foreign contacts, imported materials or trade goods of any sort. This absence suggests that even if the Yemen settlements did not exist in isolation, external stimulus is unlikely to have been a determining factor in their formation. Rather, they appear to have developed in an area which provided excellent potential for food production, and population growth. These factors do not, however, provide sufficient reason for the growth of such towns, and although the labour requirement of terraced agriculture or defence may also have contributed to the nucleation of population into such settlements, the precise reasons for their growth remains to be determined.

Acknowledgements. We wish to thank the National Science Foundation, the National Geographic Society, the American Institute for Yemeni Studies, the Oriental Institute and various private donors for funding. Thanks also go to Eleanor Barbanes for the final versions of the inked drawings; to Dr Yusuf Abdullah and colleagues at the General Organization of Antiquities and Museums, San'a, for advice during fieldwork; and two anonymous reviewers for comments on the first draft of the manuscript.

References

BRETON, J.-F. & C. DARLES. 1994. Hagar Surban 1 et 2: villages du Gabal an-Nisiyim, in N. Nebes (ed.), Arabia Felix: Beitrage zur Sprache und Kultur des Vorislamischen Arabien: 46-61. Wiesbaden: Harrassowitz.

BLAKELY, J., C. VITALIANO & R. BRINKMAN. 1996. Bronze Age remains at Kawlah al-Lajamah, Yemen Arab Republic, in W. Overstreet & J. Blakely (ed.), Environmental research in support of archaeological investigations in the Yemen Arab Republic, 1982-1987: 309-33. Washington (DC): American Foundation for the Study of Man.

BRUNSWIG, R.H. 1989. Cultural history, environment and economy as seen from an Umm an-Nar settlement. Evidence from test excavations at Bat, Oman, 1977/78, Journal of Oman Studies 10: 9-50.

CLEUZIOU, S. 1982. Hili and the beginnings of oasis life in E Arabia, Proceedings of the Seminar for Arabian Studies 12: 15-22.

CLEUZIOU, S., M.-L. INIZAN & B. MARCOLONGO. 1992. Le peuplement pre- et protohistorique du systeme fluviatile fossile du Jawf-Hadramawt au Yemen, Paleorient 18(2): 5-29.

COSTANTINI, L. 1990. Ecology and farming of the protohistoric communities in the central Yemeni highlands, in de Maigret (ed.): 187-204.

CRAWFORD, H., R. KILLICK & J. MOON. 1997. The Dilmun temple at Saar, Bahrain and its archaeological inheritance. London: Kegan Paul.

dE MAIGRET, A. 1990. The Bronze Age culture of Hawlan alTiyal and al-Hada. Rome: IsMEO.

EDENS, C. 1999. The Bronze Age of highland Yemen: chronological and spatial variability of pottery and settlement, Paleorient 25.2: 105-28.

FRIFELT, K. 1976. The evidence of a third millennium BC town in Oman, Journal of Oman Studies 2: 57-74.

HOJLUND, F. 1989. The formation of the Dilmun State and the Amorite Tribes, Proceedings of the Seminar for Arabian Studies 19: 45-59.

HOJLUND, F. & H.H. ANDERSEN. 1997. Qala'at al-Bahrain. Aarhus: Jutland Archaeological Society.

POTTS, D.T. 1990. The Arabian Gulf in antiquity. Oxford: Clarendon Press. 1991 Further excavations at Tell Abraq: the 1990 season. Copenhagen: Munksgaard.

STEFFEN, HANS. 1979. Population Geography of the Yemen Arab Republic. Wiesbaden: Dr Ludwig Reichert. Beihefte zum Tubinger Atlas des Vorderen Orients.

VOGT, B. & A. SEDOV. 1998. The Sabir culture and coastal Yemen during the second millennium BC -- the present state of discussion, Proceedings of the Seminar for Arabian Studies 28: 261-70.

WILKINSON, T.J., C. EDENS & M. GIBSON. 1997. The archaeology of the Yemen high plains: a preliminary chronology, Arabian Archaeology and Epigraphy 8: 99-142.

WILKINSON, T.J. & C. EDENS. 1999. Survey and excavation in the central highlands of Yemen: results of the Dhamar Survey Project, 1996 and 1998, Arabian Archaeology and Epigraphy 10: 1-33.

C. EDENS, T.J. WILKINSON & G. BARRATT, Edens, American Institute of Yemeni Studies, PO Box 2658, San'a, Republic of Yemen. aiysyem@y.net.ye Wilkinson, Oriental Institute, University of Chicago, 1155 East 58th Street, Chicago IL 60637, USA. t-wilkinson@uchicago.edu Barratt, Department of Ancient History and Archaeology, University of Birmingham, PO Box 363, Birmingham B15 2TT, England. Glynn@barrat.demon.co.uk

Received 29 April 2000, accepted 15 May 2000, revised 8 August 2000.