Abstract: The archaeological sites of Iraq, precious for their bearing on human history, became especially vulnerable to looters during two wars. Much of the looting evidence has been anecdotal up to now, but here satellite imagery has been employed to show which sites were looted and when. Sites of all sizes from late Uruk to early Islamic were targeted for their high value artifacts, particularly just before and after the 2003 invasion. The author comments that the “total area looted was many times greater than all the archaeological investigations ever conducted in southern Iraq and must have yielded tablets, coins, cylinder seals, statues, terracottas, bronzes and other objects in the hundreds of thousands.”
As the birthplace of the world’s first cities and many aspects of complex society, present-day southern Iraq is commonly known as the “cradle of civilization”. The large number of archaeological sites and vast quantities of cuneiform tablets and artifacts discovered in this region offer a unique historical record of these early civilizations.
During most of the twentieth century, this vast archaeological heritage was relatively well protected. This situation changed, however, immediately after the first Gulf War (2 August 1990 – 28 February 1991) and during the subsequent economic embargo of Iraq, which ended foreign-sponsored archaeological work in southern Iraq and dramatically reduced funding and personnel of the Iraq Department of Antiquities, thus leaving all but a few Mesopotamian archaeological sites virtually unprotected.
During the 1990s, sporadic reports and evidence of site looting in the south reached archaeological community, but only began to receive greater attention outside the academic community following the invasion of Iraq during the spring of 2003. Three helicopter tours conducted between May 2003 and January 2004 of major sites such as Umma, Umm al-Aqarib, Isin, Tell Schmid, Bad Tibira, Zabalam, Abbas al-Kurdi and Mashkan-shapir confirmed that significant damage had occurred.
Separate visits in the Nasariyah area to sites such as Nippur, Larsa, Adab, and Umm el-Hafriyat revealed still more damage (Gibson 2003; Lawler 2003; Figure 1). Yet these site visits and accompanying photos, which circulated throughout the media left many questions unanswered. Was looting concentrated at the largest sites, or did looting occur at smaller sites as well? How did the pattern of looting relate to the population density across the region? Did looters target sites of particular date range in the hopes of discovering particular types of artifacts? Finally, what effect (if any) did the 2003 invasion and its aftermath have on the looting of archaeological sites in southern Iraq?
Answers to such questions are possible by analyzing data derived from high resolution commercial satellite images taken both before and after the 2003 invasion by the DigitalGlobe Corporation (DGC). For this study, data from more than 1,900 sites were analyzed to determine how their location, their size, and the date of their dominant surface assemblage correspond to the evidence of looting visible in the satellite photos. By comparing imagery of the same coordinates taken both before and after the March/April, 2003 invasion, limited data were developed with respect to the timeframes in which looting occurred. The study focused on the geographic area included within three archaeological surveys:
- the Robert McC. Adams survey of the Nippur area (Adams, 1981);
- the Henry Wright survey conducted in the vicinity of Eridu (Wright, 1981); and
- the Robert McC. Adams and Hans Nissen survey around Uruk (Adams, R. McC. & H.J. Nissen. 1972.)
This area was chosen for two reasons: (1) all of the sites with anecdotal reports of significant looting lay within this region; and (2) sixty-centimeter resolution Quickbird satellite images of this same area had already been captured by DGC. The majority of images used in this study were captured either immediately before the March/April, 2003 invasion or during the following summer. Images were captured only sporadically in 2002 and after 2004. All imagery used in this study was captured by DGC later than the spring of 2006; therefore, the latest image available for this study dates to early in 2006.
A total of 9,729 square kilometers of imagery was examined, within which 0.87 percent was occupied by known archaeological sites. The Adams (1981), Wright (1981) and Adams/Nissen (1972) surveys recorded data on 1,838 sites; 95.7 percent of these (i.e., 1,760 sites within the Nippur, Eridu and Uruk surveys) have been located within the DGC imagery and form the database for this study. More than one third of the 1,838 sites have been captured in multiple images taken at least one month apart. Only rarely does imagery for any one site date more than one year apart.
We also have imagery for part of the area surveyed by Abdel-Amir Hamdani, the archaeological inspector for Dhi Qar Governorate, adding another 189 sites to our database for a total of 1,949 sites. A small number of sites surveyed but not included in the final study were apparently destroyed by modern construction. Sites in the Eridu Basin, located beyond the range of DigitalGlobe’s coverage. are not included in this study.
Locating the sites and identifying looting
Analysis began by locating each site recorded in the Nippur, Eridu and Uruk surveys on the satellite photographs. Since the surveys took place before the existence of GPS, identification in the satellite photos was performed through visual analysis. Larger sites, identified by virtue of their size, shape and location relative to known landscape features, served as anchors for locating the smaller sites in their vicinity.
Sites located in the deep desert, imaged only during the summer of 2003, were sometimes obscured by blowing sand, making them difficult to pick out – unless, of course, they happen to have been looted. Some low sites had been ploughed over and some winter shots were obscured by clouds. Despite these challenges, at least 95 per cent of the published sites, were located in the satellite images. Those sites whose locations were not identified either lay within dune fields (with no visible evidence of looting) or lay in areas that has already been ploughed.
Analysis revealed that looting holes vary in their density, distribution and visual traces. Some sites had only a single hole, others had thousands of holes. At the smaller sites, the holes tended to vary in size and shape. At the larger sites, the holes were more regular. At certain sites, looting holes with sharp edges casting deep shadows suggested recent activity. At other sites, erosion had softened their edges and partially filled in the holes, suggesting that the holes were old.
The February 2003 and December 2003 images of Tell al-Uolwiyat in Dhi Qar province in Figure 2, for example, show evidence of fresh looting in the latter image, based on the sharp edges of the holes and the dark shadows within them. By contrast, many of the same holes seen in earlier imagery are difficult or impossible to make out in the August 2003 and September 2005 imagery, with only the largest holes still casting deep shadows. Both weather and light conditions varied from image to image. We therefore had to rely on a visual estimate when examining evidence of looting, taking into consideration the size of the site, the spacing of looting holes in different areas and their spatial distribution.
After our initial analysis, we digitized the margins of each site to arrive at an accurate size for each site, more accurate than the estimate available from survey publications. During this process we made new estimates of the amount of looting at each site and compared the revised estimated with those recorded earlier. Apart from those sites examined at the outset of the project (where corrections needed to be made), the two estimates were generally consistent. Any discrepancies were within 10 percent of each other. All such estimates were made by the author. It must be stressed that, given the source material, these estimates reflect only the surface area covered by looting holes, not their depth.
Measuring the circumference of a site was very easy in some instances. In other cases, in which the boundary of the site have been encroached upon by recent agriculture or fades into the surrounding landscape, we used the dimensions of the sites recorded in site survey publications as a guide. In certain instances, especially where looting is present, it became clear that some sites were actually larger than had been recorded at the time of the ground survey.
Accurate evaluations of site size made possible by satellite imagery allowed us to divide the sites into eight groups of site sizes, as seen in Table 1. An accurate record of site size could also be combined with the estimate of how much of the site had been looted to develop an estimate of the area of site surface destroyed for each looted site.
The date of each site’s last fluorescence, derived from the publications of the surveys, was used as a means of predicting the dominant surface deposit accessible to looters. Additional data from this aspect of the project are found in the published version of this study (Antiquity, op. cit.: 127-129).
Looting trends and modern population
Modern population density in southern Iraq peaks near the Tigris, Euphrates and Gharraf rivers. Most of the archaeological data analyzed in this study come from areas that were well-watered in antiquity but were mostly desert during the 1960s and 1970s, the exception being sites from the Dhi Qar survey, which are located well within the modern irrigated zone.
The most intensely looted sites, determined by the percentage of a given site in a given area covered by looting holes, are in the central part of our study area, as shown in Figure 3a. Consequently, the areas to the north, south, east and to a lesser extent west, the data show that fewer sites have suffered from looting and the degree of looting has been less severe.
Intense looting was found to be most common close to the boundaries between the settled area and the desert, a pattern seen even more clearly in the location of those sites with the most looted hectares, the large heavily looted sites shown in Figure 3b. It seems reasonable to conclude that those sites selected for intense activity were far enough away from modern settlement to allow looting to occur unmolested, yet were close enough to dense populations to attract a large labor force.
Archaeological periods targeted by the looters
A key question for this study is whether other factors, apart from site size and location, determined which sites would be looted in southern Iraq. Is the looting of archaeological sites in this area random, or are sites selected because they have accessible material from periods especially likely to yield marketable antiquities? Did the period of occupation at smaller sites affect the degree to which smaller sites were looted? To answer these questions, we looked only at the sites recorded in the Nippur, Uruk and Eridu survey areas where an assessment of the dominant surface assemblage can be made.
Temporal data for the Uruk and Nippur survey sites is provided in a consistent manner by Adams (1981). Wright (1981), however, used different criteria, with the result that he grouped some periods together which were distinguished by Adams. This results in an apparent absence of sites dating to Early Uruk, Middle Uruk, Achaemenid, Parthian, Middle and Late Islamic periods for the Eridu Basin. Lagash has been included in this analysis even though it was only visited during the Dhi Qar survey since the intense surface survey conducted by Elizabeth Carter, et al. (1989-90) allows us to date its predominant surface collection to the later Early Dynastic period.
Not unexpectedly, early sites dating to the Ubaid and the earlier Uruk periods, listed in Table 2 and Figure 4 [available in the published version of this study (Antiquity, op. cit.: 131)], which lack the cylinder seals, tablets, statuary, coins and other elaborate objects which seem to be of interest to collectors, show relatively low levels of looting. This begins to change at sites dating from the Late Uruk and Jemdet-Nasr periods, where the sites between 3 and 9 hectares (7.4 and 22.24 acres) in size received the greatest attention, along with sites dating to Early Dynastic I period. The intensity of looting increased even more at sites dating to the later Early Dynastic and especially Akkadian sites. All but one of the six sites with Akkadian surface material showed evidence for recent looting. But the value accorded to Akkadian artefacts is nothing new since many of these sites had already suffered the attention of looters as long as four decades ago (Adams 1981: 276, 284; Adams & Nissen 1972: 224).
Summarizing the data contained in the published version of this study, we found that the date of the surface assemblages of sites does indeed have a profound impact on their likelihood of being looted. Altogether, the data suggest that looters had considerable prior knowledge of the types of artifacts they would find at individual sites and exercised considerable selectivity in which sites would be targeted.
The chronology of looting
Of the 1,949 sites examined in this study, only 743 can be seen in more than one image. And of the 743 sites imaged on more than one occasion, only 213 were judged to have been looted, representing 26 percent of all looted sites identified in this study. Our attempt to understand the timeframe in which looting occurred was therefore based on a more limited sample than was available for the rest of the analysis presented here. Among the 213 sites used in this part of the research, some sites were imaged more than a month apart on as many as five separate times. We therefore have a total of 348 pairs of images taken at different times at individual sites.
As noted above, DGC Quickbird imagery provides sufficient detail to allow fresh looting to be distinguished from older activity on the basis of the sharpness of the sides of the holes. For the earliest image of each site we therefore recorded whether or not the site had been looted and, if so, whether these traces were old or recent. For subsequent images, we recorded any visible changes, again distinguishing between fresh and older looting. The imagery has sufficient clarity that it is possible to compare pairs of images hole by hole to determine whether any new activity had taken place. Where older holes were re-excavated, which has certainly been documented by journalists who have visited the sites (Joanne Farchakh-Bajjaly, personal communication), recent activity is often visible in the form of freshly dug earth which is darker than the rest of the surface of the site.
Our data indicate that the mass of small to medium sized sites in this 243-site sample, which lay unguarded in the countryside, were treated differently from the large, well-known sites, all of which had site guards, some more effective than others.
The many small and medium sized sites in our 243-site sample show evidence of intense looting immediately before the war, suggesting that the threat of hostilities, and presumably the mistaken expectation of increased security thereafter, stimulated an unprecedented level of activity. A subset of this group, 114 looted sites, had been imaged by DigitalGlobe multiple times and were imaged in February 2003. Eighty-five per cent of these 114 small to medium sized sites showed evidence of fresh looting. In the immediate post-war period, by contrast, most of these sites show little change and the holes that were visible in the earlier imagery have been eroded. Yet by the summer of 2003, looting patterns at these particular sites appeared to resume in a more focused way.
Among this sample — consisting of 114 smaller sites that had been imaged by DGC multiple times, including immediately before the March/April 2003 invasion – it appears that as the summer of 2003 progressed, looting resumed at approximately one-third of these sites, all of which had been attacked before the war. Some sites within this subset show evidence for intense looting during the autumn of 2003, but this seems to have died down again later. The rare instances in which we have multiple images of looted sites beginning by early 2003 and continuing into 2004-2006 within this subset suggest that there was little looting at these smaller sites during the last part of our time sample (2004-06). It should be noted, however, that most of our imagery from the 2004-06 timeframe comes from Dhi Qar province, which is the one place where extensive efforts to limit the looting were conducted by the Italian Carabinieri.
These observations do not apply to the major sites, where we can bring to bear resources other than DGC imagery. Many of these sites were pictured between 2000 and 2001 in archaeological site protection planning maps provided by the Coalition Provisional Authority, Ministry of Culture, and featured in the helicopter photographs taken by John Russell and the Italian Carabinieri between the autumn of 2003 and January 2004. The major sites fall into three categories: (1) those that had been excavated for decades by foreign archaeologists; (2) those left unattended except for a few site guards; and (3) those excavated year round between 1999 and 2002 for their protection. In the latter instance, the Department of Antiquities sent permanent teams to work at some of those sites most badly looted (e.g., Umma) during the embargo period, on the theory that the presence of archaeologists at sites not only discourages looters but also provides legitimate employment for the local workforce.
It is perhaps not surprising that among the major sites excavated by foreign expeditions, Nippur and Uruk were spared the pre- and post-invasion looting, because of their long history of employing local residents. Only minor damage was found at Nippur and no damage at all at Uruk. By contrast, Larsa, a site excavated for decades by the French but whose guard had been murdered during the 1991 uprising, was very badly damaged, as shown in Figure 1. The eastern half of this ancient city was bulldozed, and when theNational Geographic group visited the site by helicopter in May 2003, there was ample evidence of recent looting on the western half of the site (Lawler 2003). No other site in our sample has anything like this evidence for the use of earth moving machinery for looting, although we do encounter occasional backhoe cuts, and at some smaller sites in the eastern portion of the Dhi Qar survey area machinery designed to stabilise dunes seems also to have been used to aid the looters. Additional data from this aspect of the project are found in the published version of this study (Antiquity, op. cit.: 127-129).
The damage to Isin, a site excavated by the Germans, has been widely reported. The UNESCO group who visited the site in June 2003 reported that looters were tunnelling down to reach lower levels (M. Van Ess, pers. comm.). This deep looting can be identified in the DGC imagery both by the larger size and blackness of the holes and by the darker color of the excavated soil. Isin is the only site where this was evident. DGC image capture dates confirm that the deep looting at Isin seems to have commenced in February 2003 and accelerated immediately following the war.
Additional data from this aspect of the project are found in the published version of this study (Antiquity, op. cit.: 127-129).
The evidence indicates that site looting in Iraq has been pervasive. Archaeological sites of all sizes and periods have been affected in the area covered by this study, with the focus of looting activity tending to favor larger sites and those dating to periods most likely to generate cylinder seals, cuneiform tablets, and, to a lesser extent, early coins. The spatial distribution of the looting suggests that many interesting sites in central Babylonia and the Eridu Basin still have intact surfaces available for investigation. If order can be re-established, a large-scale, internationally aided, rescue programme might still be able to recover useful information from the sites in the heartland of ancient Sumer.
The evidence suggests that looting of the mass of small and medium-sized sites within the examined area has died down since 2003, perhaps because the market for Mesopotamian portable antiquities has finally been saturated. This should not be surprising, given that the total area of intensive looting adds up to 15.75 square kilometers – an area many times greater than all archaeological investigations ever conducted in southern Iraq. It also remains to be seen whether the illicit market can absorb this quantity of material. If it cannot, the archaeological and Assyriological communities will need to develop a rescue programme designed to recover, conserve and eventually return these materials to Iraq.
Finally, even though contextual information has been irretrievably lost for those objects that have been exported from Iraq, even more material must still lie within the spoil heaps left next to the looting holes at the sites, which may contain potsherds and broken objects. A detailed study of the spatial distribution of the material from these dumps could recover some information from this disaster. This research should include both large cities and small sites that were heavily looted. The very fact smaller sites attracted the attention of looters suggests that they were rich in artifacts. Their investigation could redress the excessive emphasis on large sites that has characterised Mesopotamian archaeology to this day.
Additional conclusions are found in the published version of this study (Antiquity, op. cit.: 127-129).
The DigitalGlobe Corporation Quickbird imagery was purchased using funds provided by the National Geographic Society, the National Endowment for the Humanities, the U.S. State Department, the Joukowsky Family Foundation, the National Science Foundation and private donations.
Hussein, L.M., 1999. Excavations in Tell Harmal: spring 1997.Sumer, 50: 61.
Adams, Robert McC., 1981. Heartland of Cities, Chicago: University of Chicago Press.
Adams, R. McC. & H.J. Nissen, 1972. Uruk Countryside, Chicago: University of Chicago Press.
Malko, H., 2006. A comparison of the social and economic organization of cities and small settlements in Mesopotamia. Unpublished MA diss., Stony Brook University.
Carter, E., J. Black, & J. Kenoyer, 1989. A surface survey of Lagash, al-Hiba, 1984: shell artifacts from Lagash, al-Hiba; inscribed objects from Lagash, al-Hiba and Girsu. Sumer, 66: 60-83.
Pollock, S., 1999. Ancient Mesopotamia, Cambridge: Cambridge University Press.
Lawler, A., 2003. Beyond the looting. National Geographic, (October): 58-75.
Postgate, J.N., 1991. Early Mesopotamia, London: Routledge.
Baqir, T., 1948. Excavations at Harmal. Sumer, 4: 137-39.
Hussein, L.M., 2001. Excavations in Tell Harmal: fall 1998.Sumer, 51: 114-22.
Gibson, McG., 2003. From the prevention measures to the fact-finding mission. Museum International, 55(3-4): 108-18.
Kepinski-Lecomte, C., 1992. Haradum I. Paris: Recherche sur les Civilisations,
Mustafa, M.A., 1949. Sounding at Tell Dhibai. Sumer, 5: 173-86.
Al- Gailani, L., 1965. Tell edh-Dhibai, Sumer, 21: 33-40.
Baqir, T., 1946. Tell Harmal: preliminary report. Sumer, 2: 22-30.
Hamoodi, K.K., 1990. The excavations at Tell Dhibai. Sumer, 46: 91.
Wright, H.T., 1981. The southern margins of Sumer: archaeological survey in the area of Eridu and Ur, Appendix in R. McC. Adams,Heartland of Cities: 295-346, Chicago: University of Chicago Press.