geophysical diffraction tomography: new views on the shiqmim prehistoric subterranean village site...

$: Geophysical diffraction tomography: New views on the shiqmim prehistoric subterranean village site (israel)$
Geophysical Diffraction Tomography: New Views on the Shiqmim Prehistoric Subterranean Village Site (Israel)

Alan J. Witten* Energy Division, Oak Ridge National Laboratory, PO Box 2008, Oak Ridge, Tennessee 37831 -6200

Thomas E. Levy Department of Anthropology, University of California, San Diego, 9500 Gilman Drive, La Jolla, California 92093-01 01

James Ursic U S . Environmental Protection Agency Region 5 , 77 West Jackson Boulevard, Mail Stop HSRLT-SJ, Chicago, Illinois 60604

Paul White Dynamic Graphics, Inc., 1015 Atlantic Avenue, Alameda, California 94501 -1154

Shiqmim is a late prehistoric farming village in the Beersheva river valley in Israel’s northern Negev desert dating to the late 5th-early 4th millennia B.C. Along with surface architecture, archaeologists discovered a sequence of large subterranean rooms interconnected by narrow tunnels. This discovery, found in the northern Negev, as well as similar subterranean features and dating back to the same period, has created considerable controversy as to the functions of this innovative architecture. To a large extent, this controversy results from conflicting hypotheses based on the small sample of subterranean Negev architectures. A fundamental task in the resolution of these problems is the location and investigation of similar features within this region. This is not a simple task because these subterranean features can be as much as 7 m deep in hard-packed soil and are usually completely filled with fine sediment as a result of depositional forces acting over the 6 millennium since their abandonment. To help quantify the spatial extent of subterranean features at Shiqmim and identify particularly promising locations for excavation, an emerging geophysical method, known as geophysical diffraction tomography, was applied. This method is based on the propagation of acoustic waves and employs the mathematical concepts of optical holography to reconstruct quantitative, high-resolution images of the subsurface environment. Two geophysical imaging field studies were performed at Shiqmim, and the results of these studies indicate that subterranean architecture is much more pervasive at this site than pre-

* Permanent address: School of Geology and Geophysics, University of Oklahoma, 810 Energy Center, Norman, OK 73019-0628.

Geoarchaeology: An International Journal, Vol. 10, No. 2, 97-118 (1995) 0 1995 by John Wiley & Sons, Inc. CCC 0883-6353l95lO20097-22

WITTEN ET AL.

viously thought. The new geophysical studies help answer social questions and highlight the contribution of geoarchaeology to anthropological reconstructions of the past. This article discusses geophysical diffraction tomography and the first attempt to apply it in an archaeological context. 0 1995 John Wiley & Sons, Inc.

INTRODUCTION Geophysical Applications in Archaeology

There has been a rapid increase in the application of geophysical methods at archaeological sites over the past decade. Many of these applications have involved qualitative techniques, such as electromagnetic induction, metal detection, or magnetometry, to locate metallic or ferrous objects. Magnetometry, in particular, has been used (Weymouth, 1986) because it provides an indication of spatial variations in magnetic susceptibility or intrinsic magnetization that, in turn, can be used to differentiate imported or fired stones and to identify areas of disturbance to paleomagnetism as a result of past excavation. These geophysical techniques are qualitative in that they cannot quantify size and depth of anomalous features. This limitation is overcome, to some extent, in ground penetrating radar (GPR) where, by virtue of the introduction of a measurable time to target depth, a third spatial dimension can be obtained. With this additional information, it is possible to “unravel” vertical structure; for this reason, GPR has been used in archaeology to locate and identify cham- bers (Mizrachi, 1992), tunnels, burials, and other features.

The application of GPR typically involves the movement of a pair of antennas along a line on the ground surface, producing a two-dimensional “imagelike” display of a vertical cross section with the horizontal scale being distance and the vertical scale being time, which can be equated to depth. The fact that the acquired GPR data are “out of focus’’ images is exacerbated by the electromag- netically noisy subsurface environment in all but the most ideal geologies, which makes skilled interpretation of GPR imperative. The presence of conduc- tive clayey or salty soils limits the depth of penetration, which, in turn, limits the sites suitable for GPR studies.

Historically, seismic (acoustic) methods have been the mainstay of geophysical exploration. Like GPR, these are wave-based methods, but, unlike GPR, they do not typically suffer limitations associated with depth of penetration. There has been considerable development in seismic signal processing to help reduce the level of insight required in data interpretation. Although these classical techniques were developed for the identification of deep, large-scale geologic structures associated with resource exploration, geophysicists have been pursuing higher resolution and more quantitative techniques for shallow applications. The most notable of these is geotomography, which, in its earliest embodiments, applied the straight-ray methods employed in CT scanners of diagnostic medicine to acoustic wave propagation. Devaney (1984) recognized that, unlike x-rays, the longer wavelengths employed in geophysical exploration using acoustic (as well as radar) waves do not travel in straight lines and

9a VOL. 10, NO. 2

GEOPHYSICAL DIFFRACTION TOMOGRAPHY

that the interactions of these waves with subsurface inhomogeneities produce a redistribution of wave amplitude and phase known as diffraction. For this reason, Devaney proposed a geophysical imaging procedure, based on the concept of structure determination in holography (Wolf, 1969), that he called geophysical diffraction tomography (GDT). Since this early work of Devaney, GDT has been refined (Witten and Molyneux, 1992), incorporated into field instrumentation (Witten et al., 1992a), and applied to problems such as the location and identification of buried waste (Witten and King, 19901, imaging the skeletal remains of a supergiant sauropod dinosaur (Witten et al., 1992b) and detecting tunnels in the Korean demilitarized zone (Witten, 1991).

Enigmatic Prehistoric Subterranean Architecture: Excavations in Israel’s Northern Negev Desert

Until the mid-l970s, Perrot’s (1955; 1984) excavations at Abu Matar and Bir es-Safadi around the city of Beersheva had provided the foundation for understanding the nature of early 4th millennium societies in the southern Levant. Perrot’s discoveries, the underground room and tunnel complexes dating to the Chalcolithic period (ca. 4500-3500 B.C.E.), have been described as human“ant farms” or “prairie dog towns” (Levy et al., 1991; Levy and Witten, in press). This unusual architecture a t the Beersheva sites was taken as evidence of an egalitarian “troglodyte” community living along the banks of the Beersheva valley with little evidence of social differentiation (cf. Kenyon, 1985).

In summarizing the excavations at the Beersheva sites, Perrot (1984) outlines the developmental sequence as consisting of three main stages: (1) the earliest phase consists of linked subterranean room and tunnel complexes; (2) the middle phase reflects the widespread use of semisubterranean architecture; and (3) the late phase is characterized by an above ground village consisting of rectilinear buildings made of mud bricks. Perrot explained the presence of subterranean architecture in the Beersheva valley as an innovative answer on the part of this Chalcolithic culture to the problem of living in a hot arid environment. Upon reexamination of plans and sections from the Beersheva sites, Gilead (1987) refuted Perrot’s developmental model and suggested that the Chalcolithic occupation in the Beersheva valley was of a very short, single period duration. Gilead (1987) suggests that the subterranean room and tunnel complexes discovered by Perrot were in fact contemporary with the above-ground open-air villages and that they were used primarily for storage. Thus, during the 1980s, two conflicting models existed concerning the development of the Beersheva Chalcolithic culture and the function of the room and tunnel complexes. The only issue that these two scholars agreed on was the perceived egalitarian nature of the Chalcolithic cultures of the southern Levant.

GEOARCHAEOLOGY: AN INTERNATIONAL JOURNAL 99

WITTEN ET AL.

Shiqmim: New Excavations in the Beersheva Valley and New Models The discovery of Shiqmim, a pristine Chalcolithic settlement center un-

touched by modern developers, in the late 1970s by Levy and Alon (1983), has provided a unique opportunity to reinvestigate many of the developmental and social issues outlined above. Long-term interdisciplinary archaeological excavations were initiated at Shiqmim in 1979 and were completed in the fall of 1993. Located some 16 km downstream and west of the Beersheva sites, Shiqmim provides a perfect open-air laboratory for studying the questions concerning social and economic change in the Beersheva culture. The Phase I excavations at the site (1979,1982-1984) demonstrated the changing environment that faced these early farmers, particularly moister conditions during the beginning of the Chalcolithic period as well as the intricacies of the earliest floodwater farming in this region (Goldberg and Rosen, 1987; Kislev, 1987; Rosen, 1987). These excavations also showed the complexities and nature of fourth millennium domestic households in this region (Levy and Holl, 19881, the technological advances of the earliest Levantine metal industries (Shalev and Northover, 19871, and some of the social, economic, and ritual functions of an early chiefdom center (Levy, 1986; Levy and Holl, 1988).

The Phase I excavations at the site revealed one of the earliest planned village sites in western Palestine and important new information concerning the physical layout of a Chalcolithic chiefdom center. Remarkably, after 4 years of extensive excavations at Shiqmim, not one subterranean feature came to light. Although the Phase I excavations established the presence of an extensive network of large (ca. 5 X 12 m) and small (3 x 6 m) buildings interconnected by alleys and courtyards, no evidence for centralized food storage was found. As many anthropologists have shown (Renfrew, 1982), the ability to control and provision food storage is a hallmark of early complex societies, and the apparent absence of such facilities at Shiqmim highlighted the need for additional exploration of the site. Given the widespread occurrence of subterranean architecture found at sites upstream around the city of Beer- sheva, Phase I1 (1987-1989, 1993) investigations for Shiqmim were planned in 1987 with the goal of examining the entire issue of subterranean architecture in the Beersheva valley. These excavations at Shiqmim represent the first time this issue was systematically investigated on a large scale since Perrot’s initial excavations in the 1950s.

Testing the Developmental Models of a Prehistoric Culture Deep soundings were initiated at Shiqmim in 1987. A principal objective of

the new work was to obtain clear stratigraphic evidence and material suitable for radiocarbon dating to test the Beersheva culture developmental models resulting from Perrot’s original field work. The process began with the excavation of two long trenches, each over 60 m long and 2.5 m wide, across the western sector where Phase I excavations and probes produced extensive evi-

100 VOL. 10, NO. 2


dence of Chalcolithic building activities (Figure 5a). These trenches were cut down to virgin soil and reached depths of over 6 m, proving that Shiqmim had the deepest stratigraphic sequence of any of the Beersheva sites. When the sections of the easternmost trench were scraped and cleaned, the outlines of what appeared to be a subterranean storage facility with a tunnel leading from the surface were identified for the first time.

The appearance of what seemed to be large-scale subterranean architecture in the walls of the deep trenches motivated careful deep soundings around the northern aspects of the trenches. During the 1988 field season, the first conclusive evidence was found for the general outlines of Perrot’s developmental model positing a pioneer phase of subterranean occupation, followed by a later open-air village (Levy et al., 1994). By 1989, an even larger exposure was opened, extending over an area of 150 m2. In excavations from the ground surface down to a depth of about 60 cm, ash pits were found, with widespread evidence of metal work (primarily slags). These pits averaged less than 1 m in diameter and 50 cm in depth. However, one of these “pits” had no bottom and extended through the hard packed loessial sediments of the site, reaching four meters below the surface (Figure 10a). This represented the first tunnel found leading from the site surface into an extensive subterranean room complex.

Subterranean Shiqmim The deep soundings at Shiqmim have revealed at least three architectural

building strata, giving conclusive support for the general stratigraphic model originally proposed by Perrot (1955,1984). Geomorphologic studies by Goldberg (1987) in the deep trenches and in the vicinity of Shiqmim illustrate that (ca. 2-3 m) gravel deposits occur below and contemporary with the earliest pioneer occupation phases (strata I11 and IV). During this phase, the Wadi Beersheva was extremely active, compelling the pioneer settlers to build along the margins of the valley away from the active wadi channel. To cope with this geomorpho- logical constraint, the pioneer Beersheva valley settlers at Shiqmim and other sites devised a new architectural adaptation-the subterranean room and tunnel complexes. The hills, composed of dense, reworked Pleistocene loessial sediments that border the Beersheva valley provided the context for the Chal- colithic excavation of these subterranean room and tunnel complexes away from the active wadi channel. The new Shiqmim environmental data indicate that these subterranean systems flourished during the wettest phase of the late 5th millennium (Goldberg, 1987; Goldberg and Rosen, 1987; Goodfriend, 1988). Instead of being an adaptation to a hot desert environment in the pioneer phase, these architectural features provided a solution to settlement in an area that was devoid of extensive flat ground suitable for a planned open-air village. However, a larger sample of subterranean architectural features was needed to investigate function and meaning of these enigmatic units. Because traditional methods of excavation alone would have been too time-consuming to thoroughly


WITTEN ET AL.

\/ \/ \/ \/ \/ SOURCES

A A A A A GROUND SURFACE

D rn

7 rn D v)

I

Figure 1. Cross-section view of the offset VSP GDT imaging concept. The source indicated by x’s and receiver positions are indicated by solid squares.

imaging concept. The source positions are

address the research issues, new geophysical methods were used to supplement the conventional methods of excavation.


Concepts Geophysical diffraction tomography is a quantitative, high-resolution tech-

nique for subsurface imaging. The procedure is based on independent arrays of wave sources and receivers deployed along coplanar lines of arbitrary orientation (Figure 1). Mathematically the procedure for GDT-based imaging is known as “backpropagation” because the wave field, as measured over the receiver array, is propagated backwards from this measurement line into the host geology to map out spatial variations in sound speed. Conceptu- ally, this procedure mimics imaging techniques in optical holography and, as illustrated in Figure 1, employs an array of sources (or source positions) uniformly spaced along a line on the ground surface and an array of uniformly spaced receivers vertically oriented in a borehole. The geometry, known as offset vertical seismic profiling (VSP), is one of the measurement geometries employed at Shiqmim. A second geometry used at Shiqmim is seismic reflection, where sources and receivers are deployed along the same line on the ground surface. Although this discussion is based on the offset

102 VOL. 10, NO. 2


VSP geometry shown in Figure 1, the GDT concept is unchanged for seismic reflection and is a multistep procedure:

1. Data for all receivers are acquired for each source position. The individual contributions for each source are coherently summed to synthesize the response to plane wave illumination. A plane wave is characterized by parallel rays and, as such, is the analog to laser illumination in holography. This step is indicated by the synthetic aperture lens in Figure 1.

2. For a particular direction of plane wave illumination and frequency, a partial image of the subsurface is reconstructed by the application of the mathematical analog of a holographic lens. While this partial image is focused, it will be overcome with image artifacts such as elongation, halos, ringing, and streaks because insufficient information is used to reconstruct the image; therefore, i t is referred to as a partial image.

3. To produce a fully reconstructed image, a number of partial images are superimposed with each partial image associated with a variation of an independent parameter not yet exploited. For offset VSP, this parameter is the plane wave view angle. The synthetic aperture lens is applied for a fixed frequency but for a range of plane wave view angles, with each view angle contributing a partial image. In the case of seismic reflection, a single view angle (straight down) is used, and separate partial images reconstructed for different frequencies, but fixed view angle, are superimposed. In either case, the result is relatively sharp images that have a minimum of image artifacts.

The quantity imaged in GDT is the so-called object function, where, for a host formation characterized by a sound speed co and inhomogeneities characterized by some spatially variable sound speed 4x1, the object function is defined by

(1)

It is evident from the above relationship that the object function is positive for features possessing sound speeds greater than background and negative for those features having sound speeds less than background. Furthermore, for the offset VSP geometry, the sound speed of imaged features can be recovered by inverting eq. (1). Sound speed cannot be recovered from the images of O(x) reconstructed from reflection-mode measurements. This is a fundamental limitation of the measurement geometry and is independent of the signal processing. The sign of the object function is preserved in the reflection geometry and its magnitude is proportional to the difference in sound speed between spatial inhomogeneities and their surroundings. Thus, reflection-mode imaging


WITTEN ET AL.

Figure 2. Illustration of the offset VSP field implementation of GDT as used at Shiqmim. Source lines were established as spokes radially outward from the borehole and images were reconstructed over vertical cross-sections below each source line.

is inherently less quantitative than offset VSP imaging, which is a transmis- sion-mode measurement. This limitation did not present a particular problem in reflection-mode imaging at Shiqmim because all features of interest could be characterized by sound speeds less than background (negatively valued object function) and easily identified by their unique size and shape.

Implementation Dedicated data acquisition/signal processing systems have been developed

for implementing GDT in both the offset VSP and reflection geometries. Both systems use a notebook color 486-based personal computer to control data acquisition, store and display acquired data, and perform all signal processing functions. For each system, the acoustic source is a sledge hammer striking an aluminum plate moved along the ground surface. The offset VSP system uses an array of 32 hydrophones (water-coupled microphones) hard wired with 20 cm spacing. The reflection system uses 32 geophones with user-selected spacing and at Shiqmim a spacing of about 30 cm was employed to achieve the horizontal resolution required to resolve the features of interest.

The procedure for offset VSP field implementation at Shiqmim is illustrated in Figure 2. A vertical borehole was developed to a depth at least as great as the deepest anticipated feature (on the order of 8-10 m deep) and a capped plastic pipe approximately 8 cm in diameter was inserted to the full depth of the borehole. The pipe was filled with water, and the hydrophone array was lowered into the borehole. A source line was established along a line on the ground surface radially outward from the borehole. The plate was sequentially

104 VOL. 10, NO. 2


Figure 3. Illustration of the reflection-mode field implementation of GDT as used at Shiqmim.

moved and struck by the hammer a t 61-cm intervals along this line. Data were acquired over the entire source line and were processed to yield a two- dimensional image of object function over a vertical cross section having a horizontal extent equal to the length of the source line (typically about 10 m) and vertical extent equal to the depth interval spanned by the hydrophones. The image reconstruction is executed in several seconds on the personal computer and can be viewed in the field as a gray-scale rendering. The process is repeated over additional source lines until the area around the borehole has been completely imaged. A new borehole can be developed in another area and the above-described sequence repeated.

The reflection-mode implementation (Figure 3) followed a procedure similar to that employed for offset VSP imaging; however, geophones were placed along a line on the ground surface, with the hammer and plate applied over this same line at points between each adjacent geophone pair. As with offset VSP imaging, a two-dimensional image can be reconstructed and displayed on the personal computer within several seconds after data acquisition. Here the image is over a vertical cross section having a horizontal extent equal to the length of the geophone line (at Shiqmim, about 10 m). The vertical extent is user defined but is limited by the time window over which data are acquired. To image to a specified depth, data must be acquired over a time sufficient to allow sound waves to propagate to a feature and be reflected back to the ground surface. Following data acquisition and signal processing along a particular line, the geophone array can be relocated. The general procedure applied at Shiqmim was to survey over equally-spaced parallel lines, as depicted in Figure 3, although spacing and orientation were occasionally modified to conform to local topography.


WITTEN ET AL.

Figure 4. Illustration of the ambiguity in feature location that can occur in a reflection measurement geometry. Any feature appearing in an assumed vertical cross section below the geophone line (outlined by a solid line) can occur within any cross sectinn defined by rotation of this cross section about the geophone line. An example of such a rotated cross section is outline by a dashed line.

As noted above, images reconstructed by GDT are two-dimensional. This introduces the assumption of target two-dimensionality when, in fact, almost all subsurface inhomogeneities are truly three-dimensional. For two-dimensional imaging to be rigorously correct, all inhomogeneities must have constant size, shape, orientation, and material properties in the direction perpendicular to the imaged cross section. Although GDT imaging algorithms are easily generalized to three dimensions, their implementation requries that sources and receivers be deployed over planes rather than lines, making the direct use of such algorithms prohibitive because of instrumentation and data acquisition costs. It is, therefore, more practical to consider two-dimensional implementa- tions and be cognizant of the underlying assumption in interpretation. For offset VSP, the two-dimensional GDT algorithm does not account for “out-of- cross-section” scatter resulting from sound passing through features not within the plane of the vertical image cross section, being refracted, and subsequently recorded by the receivers. As discussed in Witten et al. (1992b1, this is not a significant concern because the imaged cross section is essentially “anchored” on two sides: on the top by the source line and on one side by the receiver array. The situation is completely different for the reflection geometry. Here, the imaged cross section is anchored only along the surface by the superimposed sourceh-eceiver line. In fact, no cross section is defined at all; it is only assumed to be directly below this line. Thus, features occurring in some portion of a half-cylinder will be manifested within the imaged cross section, as illustrated in Figure 4. Here, a feature imaged within a vertical cross section (depicted in dark gray) can exist within a different cross section (depicted in light gray) that is rotated about the measurement line (outlined by a dashed line).

106 VOL. 10, NO. 2


FIELD STUDIES AND RESULTS The application of geophysics at Shiqmim posed some unique challenges.

Complete exploration by traditional excavation methods is impractical a t Shiq- mim because of the size of the site and because depositional forces acting over the 6 millennium since its abandonment has resulted in the complete or near complete filling of these features. The identification of “soft” depositional material with the “harder” naturally occurring site soil not only makes discovery by excavation tedious but also pushes the limits of discovery by geophysics. This is because the features of interest are relatively small (1 m diameter tunnels, rooms with characteristic dimensions of 2 or 3 m) and present a very weak acoustic or electromagnetic contrast compared to the host soil.

Two GDT field studies were performed at Shiqmim. The first, a limited feasibility study, was executed near the originally discovered room and tunnel complex in 1992, and the second, a full-scale geophysical exploration, was undertaken in 1993. Together, these studies helped identify the function of subterranean architecture in the northern Negev (Levy and Witten, in press). The 1992 study was of limited extent, employing the offset VSP system, and primarily served to establish the efficacy of GDT for characterizing subterranean Shiqmim. The 1993 imaging effort used the, then newly developed, reflection-mode system and was executed as part of a 6-week-long field season.

The 1992 Study During the 1992 field study, offset VSP imaging was used to reconstruct

images in a region just north of the surface Chalcolithic architecture (Figure 5a) that encompasses areas of known (Figure 5b) and suspected subterranean features. The first imaged cross section was below line A1 (Figure 5a) chosen in order to image a portion of a tunnel that had been partially excavated and subsequently backfilled. The image of this cross section is shown in Figure 6, where increasing values of object function are displayed as darker shades of gray. Because the subterranean rooms and tunnels are characterized by negative values of object function, the features of interest should appear in the images as white or light gray. The white area at middepth on the left side of this image appears to be a tunnel cross section. Applying eq. (l), a measured background sound speed of 500 m/s, and the reconstructed value of object function yields a sound speed within the tunnellike feature of 450 m/s. This particular cross section lies just to the east of the furthest extent of the tunnel excavation. The continuation of this tunnel, consistent with this image, was confirmed by excavation in 1993. The GDT image of the cross-section below line A2 (Figure 5a) revealed a similar feature that matched the size and location of features known from earlier excavations. This is the portion of the tunnel that was backfilled after excavation, and it is interesting that the reconstructed value of sound speed was about 400 m/s, making the sound speed of the backfill 50 m/s lower than the original depositional fill.


WITTEN ET AL.

I H I I K L M N O P Q R S T

i '7

+ + + + + + + +-------.

Figure 5. The 1992 Shiqmim study area as illustrated by (a) the site plan showing the excavated surface and subterranean features and (b) a photograph of the Sub-room 8 excavation.


Shiqmim - Line A 1

Figure 6. Gray-scale image of the vertical cross section below line A l . A tunnel is evident at left.

Having established the capacity for GDT imaging to resolve a known subterranean feature, the exploration moved to a hilltop immediately north of the previously discovered subterranean complex. A borehole was developed in the center of the region (Borehole B, Figure 5a) and cross-sectional images were reconstructed from sources located along lines B1 through B5 (Figure 5a). These five images revealed a complex structure of additional rooms and inter- connecting tunnels within this loessial hill. A particularly interesting cross- sectional image is that below line B4 (Figure 7). Here, a subterranean room is evident in the lower right of the image, and a tunnel extends from this room towards the ground surface. A horizontal tunnel appears in the upper left of the image, and the image below line B5 suggests that this tunnel terminates in a room.

Two additional cross-sectional images were obtained below lines C1 and C2 (Figure 5a). Steep slopes to the north and south of these lines constrained the imaging to an east-west direction. Although these topographic constraints removed the possibility of three-dimensional perspective, the images below lines C1 and C2 suggested the existence of two more subterranean rooms.


WITTEN ET AL.

Figure 7. GDT image of the vertical cross section below line B4. This image reveals a room (lower right) with vertical tunnel and the cross-section of a horizontal tunnel (upper left).

The 1993 Study The objectives of the 1993 field study were to demonstrate the feasibility of

reflection-mode GDT for imaging subterranean room and tunnel complexes and to establish the spatial extent of these features at Shiqmim. To meet the first objective, this field study was scheduled to coincide with a planned excavation thereby allowing the rapid verification of images. A loessial hill (Figure 8) approximately 300 m east of the main excavation (Figure 5a) was selected for imaging to obtain an indication of how pervasive subterranean features are at Shiqmim. The hillside, labelled Area X, rises about 12 m above the present day Wadi Beersheva channel. Area X was taken to be representa- tive of any of the 10 hills that comprise the Shiqmim upper village.

Fourteen geophone lines were established for imaging in Area X. The first six lines were on the top of the hill, with subsequent lines proceeding southeasterly down the hill towards the wadi. Figure 9a is a gray-scale GDT image reconstructed from one of the geophone lines on the hilltop. Because of the limitations of reflection mode imaging, the values of object function have no absolute significance; but the sign of the object function is important. Thus, the white and light gray features in Figure 9a indicate regions that are less dense than the host soil. The lighter features in this image suggest a surface access tunnel extending from the upper central portion of the image diagonally downward to a room in the lower left. Two-dimensional cross-sectional images from geophones adjacent to this one supported this hypothesis, and excavations of this area revealed this room and tunnel (Figure 9b). It was further discovered that

110 VOL. 10. NO. 2


Figure 8. Site plan for the Area X GDT study.

the room did not lie directly below the geophone line as indicated in the image but was several meters southeast of the imaged cross-section. A spatially correct interpretation of Figure 9a can be made by rotating this image about 30” with respect to the vertical (Figure 4). Although after the fact, this correction was made by measuring the difference in distance (depth) to the room on images below two geophone lines. With this measured distance, an appropriate rotation can be estimated by triangulation. Collectively, the 14 cross-sectional images obtained in Area X revealed that the hillside was honeycombed with subterranean features.

COMPUTERIZED EXCAVATIONS The cross-sectional images obtained during the 1992 and 1993 field studies

were taken one step further through the application of three-dimensional rendering and computerized excavation which previously had never been at- tempted on archaeological site data. This new application was accomplished by using a specialized, commercially available software package called Earth- Vision (Earthvision is a registered trademark of Dynamic Graphics, Inc.) which allows the three-dimensional display of semitransparent objects. Some of the features available with Earthvision are rotate, pan, zoom, and the ability to strip away either horizontal or vertical layers of the image. With the EarthVi- sion software, a spatially varying property (here, the object function) is rendered in false colors.0ne important capability of Earthvision is the option to turn off or make transparent selected colors, allowing subsurface architecture to be viewed through the host geology and, more importantly, “computer exca-


WITTEN ET AL.

Shiqmim Line DG2, Co=l600 f p s

Figure 9. Images of Shiqmim Area X as (a) the gray-scale, reflection mode GDT reconstruction of a vertical cross section showing a room and surface access tunnel and (b) a photograph of the room following excavation.

112 VOL. 10, NO. 2

GEOPHYSICAL DI FFRACTI ON TO MOG RAPHY

vated” by turning off the particular color associated with the interior of these features.

Figure 10 is an example of such an “excavation.” Figure 10a is a photograph of a tunnel in the originally discovered room and tunnel complex (Figure 5a) after traditional excavation. Figure 10b is a computer excavated tunnel in the vicinity of borehole B (Figure 5a). This rendering was obtained using Earthvision by removing the colors associated with small (negative) values of object function. A comparison of the images shown in this figure clearly shows that GDT images as rendered by Earthvision bear a striking resemblance to an actual excavated tunnel. Similar computer excavations from the 1992 image showed features consistent with those previously excavated, such as storage silos in the floors of rooms.

A second three-dimensional Earthvision rendering is given in Figure 11, which shows the entire Area X region studied in 1993. Here, the host loessial sediment has been made transparent and the soft depositional material has been removed so that only the boundaries between hard and soft soil remain. Presented in this fashion, the full extent of the suspected subterranean architecture can be appreciated.

SUBTERRANEAN SETTLEMENT At Shiqmim, GDT proved to be a powerful technique for supplementing

traditional methods of subterranean exploration. In addition to locating a tunnel entrance in Area X leading to an underground room, GDT results indicate that this loessial hill is literally honeycombed with subterranean architectural features (Figure 11). The astounding GDT images highlight a number of issues associated with the Chalcolithic subterranean features found at Shiqmim including: (1) The subterranean architecture at Shiqmim extends for over 400 m across this early village settlement center; (2) based on the sampling of Area X, it can be assumed that all of the loessial hillocks at the site are permeated with subterranean features; and (3) while the function of the subterranean features identified in geophysical surveys can be confirmed only by excavation, the widespread distribution of these features suggests that many were linked to storage and defensive needs.

Whereas Perrot (1984) suggests that the subterranean architecture went out of use when open-air village settlements thrived during the last occupation phase in the Beersheva valley and Gilead (1987) believes that these enigmatic structures were all contemporary with open-air villages, the new Shiqmim data suggests a more complicated picture. The stratigraphy and radiocarbon dates from Shiqmim indicate that the earliest settlement was indeed characterized by subterranean architecture (Levy et al., 1991). In this pioneer phase, the subterranean architecture was an adaptive response to both the natural and cultural environment. Unlike Perrot (19841, who suggests that the subterranean room complexes were an adaptation to the hot desert environment, our

CONCLUSIONS-NEW MODELS FOR UNDERSTANDING


WITTEN ET AL.

Figure 10. Comparison of (a) a traditionally excavated tunnel with (b) a computer excavated tunnel.

114 VOL. 10, NO. 2

GEOPHYSICAL DI FFRACTION TOMOG RAPHY

Figure 11. Earthvision rendering of Area X showing a honeycomb of subterranean features.

team has shown that climatic conditions during the pioneer settlement phase were more humid. The underground room complexes from this phase (Strata III-IV) were established on the margins of the valley, inside the loessial hills, because there were no extensive floodplains on which to build an open-air village. The construction of the planned village at Shiqmim (Figure 5a) oc- curred in Stratum 11, toward the end of the Chalcolithic sequence when the natural floodplain in the Beersheva valley built up.

The hypothesis that the earliest subterranean room and tunnel complexes were used for defense is based on remarkable ethnographic parallels from east Africa. In northern Tanzania, Fosbrooke (1953,1954) has shown how a number of tribes (including the Gweno, Iraqw, Chagga, and Sonjo) in the early 1950s still excavated subterranean bolt holes connected to underground room and tunnel complexes for defense. Fosbrooke’s ethnographic descriptions, plans, and photographs show that the east African examples are virtually identical to those in Chalcolithic Palestine. The presence of at least three other contemporary Chalcolithic cultures in the northern Negev (Levy, 1986) and the widespread distribution of mace heads, a clear example of hand-to-hand combat (Yadin, 19631, argue for the possibility of intergroup conflict in the Negev at this time.

In the later strata at Shiqmin (Strata I1 and I), the subterranean room and


WITTEN ET AL.

tunnel complexes take on additional functions related to large-scale storage of grain (barley and wheat). When the excavation of Subterranean Room 8 was completed in the fall of 1993, a remarkable discovery was made which links it stratigraphically with the planned Stratum I1 open-air village at Shiqmim. A well-defined, hard-packed Stratum I1 surface was found linking a large rectilinear public building with the entrance to Sub-Room 8. The entrance to Sub- Room 8 was extraordinary in that it is stone lined with quarried chalk blocks and two carefully carved chalk lentils, each over one meter in length. After passing through this entrance, a small anteroom was constructed with small storage bins. The anteroom then opened into the main chamber associated with Sub-Room 8. In the back of this underground room, a huge storage silo for grains was found. It is about 1.8 m in diameter and 2 m deep. Associated with this storage facility was a carefully excavated pit containing a large grinding stone used in the processing of grain.

When the new excavation is considered in conjunction with the radiocarbon date linking Sub-Room 8 to Stratum 11 and the widespread distribution of subterranean architecture at Shiqmim, the multiplicity of functions concerning this phenomena becomes more clear. The date from this underground room (RT-1322) yielded a determination of 5190 2 75 B.P. When calibrated, this is equivalent to a range of 4212-3829 B.C. Whereas the earliest excavations failed to produce evidence for public storage of grain at Shiqmim, new research has succeeded in identifying actual examples of public grain stores associated with the later strata. GDT images reveal the widespread nature of these storage activities and other possible functions of the subterranean architecture a t Shiqmim.

In summary, the construction of subterranean architecture was a feature of Chalcolithic settlement in the Beersheva valley from the earliest pioneer phase and throughout the sequence. Underground room and tunnel complexes were a local innovation and a response to both the natural and cultural environment of the northern Negev region. GDT has been instrumental in demonstrating the presence and widespread distribution of subterranean architecture at Shiq- mim. Of the two geometries employed at Shiqmim, offset VSP is preferred because it offers a quantification of subsurface properties. It is, however, slower to deploy than the reflection method and may present logistical problems in remote locations, steep terrain, or the harsh desert conditions of Israel's northern Negev. Although seismic methods are more time consuming to implement than some other geophysical methods, such as magnetometry, the 1993 field study established that GDT imaging can be accomplished much faster than excavation. Although GDT might be impractical for large area searches, it certainly promises to be a significant tool for focusing the excavation efforts once a site has been selected.

The authors would like to acknowledge the binational support this project received from the governments of the United States and Israel. These institutions and individuals include: The

116 VOL. 10, NO. 2


US. Department of Energy; the US. Environmental Protection Agency; Oak Ridge National Laboratory, Tennessee; Admiral James Watkins, former US. Secretary of Energy; Professor Yuval Ne’eman, former Israeli Minister of Science and Technology; the U.S. Embassy in Tel Aviv; and Professor A. Biran, director of the Nelson Glueck School of Biblical Archaeology and the Hebrew Union College Jewish Institute of Religion (HUC JIR) , Jerusalem. Financial support for the Shiq- mim excavations has generously come from the National Endowment for the Humanities, Washing- ton, D.C.; the National Geographic Society; the C. Paul Johnson Family Charitable Foundation (Chicago); and the Samuel H. Kress Foundation (New York). Field work was greatly facilitated with the help of David Alon, Israel Antiquities Authority; Yorke Rowan; Morag Kersel; the Samuel H. Kress Fellows, HUC J I R , Jerusalem; Wilson and Peggy Bechtel, Southwest Paleontology Soci- ety, Albuquerque, New Mexico; Irit Yazurski and Kibbutz Hazerim. Three-dimensional imaging could not have been performed without the generous assistance of Wayne Dickson, Dynamic Graphics, Inc., Bethesda, Maryland and the Tel Aviv sales office of Silicon Graphics, Inc.

REFERENCES Devaney, A.J. (1984). Geophysical Diffraction Tomography. IEEE Transactions on Geoscience and

Remote Sensing GE-22, 3-13. Fosbrooke, H.A. (1953). The Defensive Measures of Certain Tribes in Northeastern Tanganyika,

Part I. Tanzania Notes and Records 36, 1-6. Fosbrooke, H.A. (1954). The Defensive Measures of Certain Tribes in Northeastern Tanganyika,

Part 11-Iraqw Housing Affected by Inter-Tribal Raiding. Tanzania Notes and Records 37, 50-57.

Gilead, I. (1987). A New Look at Chalcolithic Beersheva. Biblical Archaeologist 50, 110-117. Goldberg, P. (1987). The Geology and Stratigraphy of Shiqmim. In T.E. Levy, Ed., Shiqmim I , pp.

Goldberg, P., and Rosen, A.M. (1987). Early Holocene Paleoenvironments of Israel. In T.E. Levy,

Goodfriend, G. (1988). Mid-Holocene Rainfall in the Negev Desert from 13C of Land Snail Shell

Kenyon, K. (1985). Archaeology in the Holy Land. London: Methuen. Kislev, M. (1987). Chalcolithic Plant Husbandry and Ancient Vegetation at Shiqmim. In T.E.

Levy, T.E. (1986). The Chalcolithic Period. Biblical Archaeologist 49, 83-106. Levy, T.E. and Alon, D. (1983). Chalcolithic Settlement Patterns in the Northern Negev Desert.

Current Anthropology 24, 105-107. Levy, T.E. and Holl, A. (1988). Les Priemieres Chefferies de Palestine. La Recherche 203,

1166-1 175. Levy, T.E., and Witten, A.J. (in press). New Geophysical Imaging Techniques Appliedto a Subterra-

nean Village in the Negev Desert, ca. 4500 B.C. Archaeology Magazine. Levy, T.E., Alon, D., Grigson, C., Holl, A., Goldberg, P., Rowan, Y., and Smith, P. (1991). Subterra-

nean Negev Settlement. National Geographic Research and Exploration 7, 394-413. Levy, T.E., Alon, D., Goldberg, P., Grigson, C., Smith, P., Buikstra, J.E., Holl, A., Rowan, Y.,

and Sabari, P. (1994). Protohistoric Investigations a t the Shiqmim Chalcolithic Village and Cemetery: Interim Report on the 1988 Season. In E.D. Dever, Ed., Preliminary Excavation Reports: Sardis, Paphos, Caesarea, Maritima, Shiqmim, Ain Ghazal, pp. 87-106. The Annual of the American Schools of Oriental Research Volume 51. Baltimore: American Schools of Oriental Research.

35-43. BAR International Series 356. Oxford: BAR.

Ed., Shiqmim I, pp. 23-33. BAR International Series 356. Oxford: BAR.

Organic Matter. Nature 333, 757-760.

Levy, Ed., Shiqmim I , pp. 251-279. BAR International Series 356. Oxford: BAR.

Mizrachi, Y. (1992). Mystery Circles. Biblical Archaeology Review 1 8 , 4 7 4 7 . Perrot, J . (1955). The Excavations a t Tell Abu Matar Near Beersheva. Israel Exploration Journal

Perrot, J . (1984). Structures d’Habitat Mode de Vie et Environment. Les Villages Souterrains des 5,17-40,73-84,167-189.


WITTEN ET AL.

Pasteurs de Beersheva dans le Sud #Israel, au IVe Millenaire Avant le l’Ere Chretienne. Paleorient 10, 75-96.

Renfrew, C. (1982). Socioeconomic Change in Ranked Societies. In C. Renfrew and S. Shennan, Eds., Resource and Exchange-Aspects of the Archaeology of Early European Society, pp. 1-8. London: Cambridge University Press.

Rosen, A.M. (1987). Phytolith Studies at Shiqmim. In T.E. Levy, Ed., Shiqmim I, pp. 243-249. BAR International Series 356. Oxford: BAR.

Shalev, S., and Northover, P. (1987). The Chalcolithic Metal and Metal Working from Shiqmim. In T.E. Levy, Ed., Shiqmim I, pp. 357-371. BAR International Series 356. Oxford: BAR.

Weymouth, J.W. (1986). Geophysical Methods of Archaeological Site Surveying. In M.B. Schiffer, Ed., Advances in Archaeological Method and Theory, Volume 9, pp. 311-395. Orlando: Aca- demic.

Witten, A.J. (1991). The Application of a Maximum Likelihood Estimator to Tunnel Detection. Inverse Problems 7 , L49-L55.

Witten, A.J., and King, W.C. (1990). Sounding Out Buried Waste. Civil Engineering 60, 62-64. Witten, A.J. and Molyneux, J.E. (1992). Geophysical Diffraction Tomography: Validity and Imple-

mentation. In J.B. Bednar, L.R. Lines, R.H. Stolt and A.B. Weglein, Eds., Geophysical Inversion, pp. 354-369. Philadelphia: SIAM.

Witten, A.J., Stevens, S.S., King, W.C. and Ursic, J.R. (1992a). Field Studies in Geophysical Diffraction Tomography. Inverse Problems in Scattering and Imaging 1767, 177-185.

Witten, A.J., Gillette, D.D., Sypniewski, J., and King, W.C. (1992b). Geophysical Diffraction Tomography at a Dinosaur Site. Geophysics 57, 187-195.

Wolf, E. (1969). Three-Dimensional Structure Determination of Semi-Transparent Objects from Holographic Data. Optical Communications 1, 153.

Yadin, Y. (1963). The Art of Warfare in Biblical Lands. London: Weidenfeld and Nichols.

Received September 12, 1994 Accepted for publication October 19, 1994

118 VOL. 10, NO. 2

geophysical diffraction tomography: new views on the shiqmim prehistoric subterranean village site...

Documents