Magnetic Gradient and Electrical Resistivity Tomography Surveys in Meroe, the Capital City of the Kush Kingdom, Sudan

ABSTRACT

One of the most important archaeological sites in Sudan is Meroe, capital city of the kingdom of Kush (800 bc to ad 400). The site is located in the Nile valley, about 200 km northeast of Khartoum. The most prominent feature among the ruins of Meroe is the Royal City: a stone-walled enclosure containing remains of palaces, governmental buildings and temples. The current study presents the results of integrated magnetic gradient and electrical resistivity tomography surveys in unexplored and partly explored areas in the central part of the site and in the southern part of the Royal City. The main enclosure wall, remains of sandstone and red-brick buildings, and a number of small archaeological structures have been traced on magnetic maps. The extent of buildings, identified by magnetic survey, has been complemented by information on the depth of structures provided by electrical resistivity tomography. The geophysical results have been partly verified through excavation. Copyright © 2012 John Wiley & Sons, Ltd.

Introduction

Meroe City (Figure 1a and b) is located on the east bank of the River Nile in central Sudan and is built upon the Cenomanian Shendi Formation (Figure 1b), which is predominantly a fine-grained clastic sediment deposited from a braided-meandering river (Bussert, 1993). The position of the Nile relative to Meroe City may have changed over the past 2500 years. It is assumed that the original ancient settlement in the areas close to the Nile was located on unstable alluvial islands in a braided channel (Wolf, 2004).

Figure 1. Location, geology and archaeological maps of Meroe (Capital of Kush, Sudan): (a) location of Meroe in Sudan, showing Kerma and Jebel Barkal, the former capitals of the Kushite–Napatan kingdoms; (b) geological map of the Meroe area and its surroundings (modified after Mohamed-Ali, 2007); (c) the archaeological map of the central part and the Royal City of Meroe showing locations of the geophysical surveys (modified after Shinnie and Anderson, 2004). This figure is available in colour online at wileyonlinelibrary.com/journal/arp.

The Kushite Kingdom had appeared in the Dongola Reach (Figure 1a) by the eighth century bc, centred on the Napata region (Vincentelli, 2006) with its capital being established in Jebel Barkal near the fourth cataract of the River Nile (Figure 1a). However, the royal residence (centre of government) was transferred from Napata south to Meroe as a consequence of the campaign of Psamtik against Kush in 592 bc (Török, 1997) and accordingly Meroe becoming the capital of the Kushite Meroitic Kingdom. Most prominent among the ruins of Meroe (Figure 1c) is the stone-walled enclosure containing the rubble remains of palaces and governmental buildings, several small temples and the so-called ‘Royal Bath’. Immediately to the east sprawls another walled compound enclosing the Amun Temple, a near replica of the one at Jebel Barkal (Kendall, 1982). A mixture of various construction techniques was used in building Meroe, for example, mud bricks, baked red-bricks along with stone slabs and rubble, or combined with one or two-sided brick facing (Wildung, 1996). The enclosure wall in Meroe (Figure 1c) was composed of massive stone with a thickness of approximately 3 m (Hakem, 1988). Initial excavations at Meroe were carried out by Garstang in early of the 20th century (Garstang et al., 1911; Török, 1997), then by P. L. Shinnie on behalf of University of Khartoum and University of Calgary between 1965 and 1972 (Shinnie and Bradley, 1980). The recent excavations in the areas surveyed in the current study are documented by Grzymski (2003 and 2005).

Geophysical data acquisition and processing

Geophysical prospection in Meroe, carried out for the Canadian Royal Ontario Museum expedition, commenced in 2002. Magnetic surveys were carried out in the central part of the town (Figure 1c) to the south of a road leading to the Temple of Amun, and in the southern part of the Royal City. Electrical resistivity tomography surveys were selected based on magnetic results. Magnetic surveys were carried out using a Geoscan Fluxgate Gradiometer (FM36) while the electrical resistivity tomography (ERT) was acquired using the SARIS resistivity imaging system.

Magnetic survey

The magnetic surveys were laid out in a grid pattern (20 × 20 m). The measurements for each grid were collected in parallel traverses 0.5 m apart with a traverse sampling interval of 0.25 m. The magnetic results are presented as grey tone images with negative values showing up as black and positive as white. Geoplot software was used for magnetic processing such as despiking and spatial filtering to enhance subtle patterns and weak signals and to suppress unwanted noises. Golden Software Surfer 8 and CorelDraw were used for preparation of the final magnetic images, and correlating results with the excavation.

The magnetic survey (Figure 1c) covered about 2.5 ha. Area-I (ca. 0.36 ha) was located directly to the southwest of the palace M 750, and was designed to investigate the suspected extension of M 750 (Török, 1997). Area-II covered ca. 0.87 ha targeting extensions for the partially excavated temples M712 (Grzymski, 2005) and MJE105 (Wolf, 1996); to prospect for hidden and unexcavated structures, and to test whether the mud-brick building materials were detectable. The largest magnetic survey (1.27 ha) was in area-III (Figure 1c). The target included part of a low-lying outcrop of the main enclosure wall and extended in other areas so as to define its orientation, and to prospect for new archaeological structures.

Electrical resistivity tomography

Electrical resistivity tomography (ERT) survey was carried out at six locations in Meroe City (Figure 1c) using Wenner and dipole–dipole electrode arrays with electrode spacing varying according to the objectives of the survey and the scales of the target features. Five ERT profiles (Figure 2b) were carried out in the area where the extension of palace M 750 was assumed by Török (1997). The extension indicated was explored in the current study by magnetic survey (Figure 2a). The ERT profiles (1–4) were carried out 4 m apart with an electrode spacing of 2 m along the profiles. These profiles were intended to test the applicability of resistivity tomography for estimating the vertical extensions of the detected magnetic anomalies. Profile 5 was conducted in order to resolve the target anomalies in a different direction. An electrode spacing of 0.5 m was preferred in this profile in order to resolve the anomalies more precisely. Profile 6 (Figure 3b) was carried out to estimate the unknown vertical extension of the enclosure wall, whose lateral extent and layout were defined by magnetometry.

Figure 2. Magnetic survey of area-I, southwest of palace M750 and the corresponding interpretation and verification: (a) magnetic map; (b) interpreted layout of the detected building, showing locations of the electrical resistivity tomography profiles and the test excavation; (c) photograph of the test excavation carried out by K. Grzymski at the location of the magnetic anomaly outside the discovered building. This figure is available in colour online at wileyonlinelibrary.com/journal/arp.

Figure 3. Magnetic surveys in Area-II and Area-III in Meroe: (a) magnetic map of Area-II, around temple M712 in the central part of Meroe; (b) the magnetic map of Area-III near the Royal Bath in the southern part of Royal City showing the location of electrical resistivity tomography profile 6.

The measured ERT data points were plotted as pseudosections (e.g. upper section in Figure 4) and true resistivity sections (e.g. lower section in Figure 4) were produced by the inversion processes using Res2Dinv inversion software (Geotomo, 2004). In such processes, forward modelling was used to calculate the apparent resistivities (e.g. middle section in Figure 4), and a non-linear least-square optimization technique was used for the inversion routine (DeGroot-Hedlin and Constable, 1990; Loke and Barker, 1996). The ERT processing in the current study was carried out using blocky (robust) inversion, as described by Loke et al. (2003). The advantages of blocky inversion compared with a smooth inversion technique were provided by Olayinka and Yaramanci (2000). The blocky inversion scheme was preferred here as it can produce more reliable anomalies with sharp boundaries and reduce the effect of noisy data points. The optimum results were arrived at using models where the width of the cells was set to half the unit electrode spacing. The finest mesh with four nodes between electrodes was preferred and a Jacobian matrix was recalculated each iteration so that the apparent resistivity was calculated more accurately. The above options provide accurate results for our case study where a large resistivity contrast was expected. A line search method was used to reduce RMS error and the finite difference method was used for a faster inversion as the data set did not contain topography. CorelDraw was also used here to correlate between different profiles displaying the archaeological interpretation.

Figure 4. Inversion example of electrical resistivity tomography data using Res2Din software; the upper section is the measured resistivity data, the middle section is the calculated resistivity pseudosection; the lower section is the inverse resistivity model. This figure is available in colour online at wileyonlinelibrary.com/journal/arp.

Results and interpretation

Magnetic survey

The magnetic images show a high level of noise, probably due to hidden baked brick fragments and metallic detritus from the previous excavations. The amount, strength and character of the noise in the processed images did not mask primary archaeological anomalies in the magnetic images.

The magnetic results in area-I (Figure 2a) show regular, linear intersecting anomalies. These linear anomalies are associated with a sandstone walled building. The anomalies are detectable owing to the contrast between the sandstone's non-magnetic constituents and the magnetically susceptible surrounding soil. As Meroe was proposed to be located on a small Nile island (Bradley, 1982), the soil is considered to have formed in sediments deposited at the site over many years by annual River Nile floods. The imaged building measures approximately 35 m × 55 m and appears to comprise of a number of small square rooms and corridors, each room measures 5 m × 5 m. A single room outside the building indicated was selected for excavation (Figure 2).

The magnetic image of area-II (Figure 3a) included temple MJE105. A weak magnetic anomaly, visible in the northwestern part, has a relatively regular geometry and is probably associated with the unexcavated part of extension of temple MJE105. A broad noise-free linear anomaly is associated with the main road to the Amun Temple.

The magnetic image of area-III (Figure 3b) has a low noise level but lacks interesting anomalies other than those associated with a broad linear non-magnetic anomaly of almost zero value, which coincides with the known orientation of the main enclosure wall. An additional area (40 × 20 m) shows two anomalies with almost the same dimension as the enclosure wall but with different constituents and trends (Figure 3b). The first anomaly shown by the dashed arrow is about 1 m thick shows relatively clear edges, and it is interpreted accordingly as a spur off the main enclosure wall. The edges of the second anomaly shown by the solid arrow are not as sharp as the first as it comprises multidipole small anomalies and probably results from a 'baked' brick wall.

Electrical resistivity tomography

Electrical resistivity tomography profiles 1 to 4 across the magnetic anomalies in area-I (Figure 2b) were carried out using both Wenner and dipole–dipole arrays with an electrode spacing of 2 m. The inversion process for Wenner data using the robust technique converged after 10 iterations with a RMS misfit between 5.5% and 18.5%, whereas for the dipole–dipole data (also inverted using the robust inversion method) the inversion process converged after 15 iterations, with a RMS misfit ranging between 10.9% and 20.4%. The latter and the corresponding iteration number are relatively larger than that of Wenner sections because the dipole–dipole array is more sensitive to shallow subsurface heterogeneity and noise contents (Dahlin and Zhou, 2004). The interpretation of these ERT profiles was achieved in light of the previous excavation in the area (Török, 1997; Grzymski, 2005).

The inverted sections (profiles 1–4, Figure 5a) show vertically elongated rectangular-shaped anomalies of about 1.5 m vertical extension with resistivity values of ca. 40–80 Ω.m on top of a zone with much lower resistivity ranging from 2 to 40 Ω.m. The resistivity values of these anomalies lie within the standard resistivity range of sandstone (Kearey and Brooks, 1991). The excavation carried out by Grzymski (2005) confirmed the interpretation of sandstone walls (Figure 2b). The excavation showed that the sandstone walls were very soft and crumbled on touch. It also showed that the relatively low resistivities (8–25 Ω.m) are associated with humus and clayey soils formed on sediments deposited during the annual flooding of the River Nile. The boundaries of these anomalies along profiles 1 and 2 are not sharp compared with profiles 3 and profile 4. The smoothing, separation and continuity of the anomalies may be associated with the collapse of the walls inside the rooms, in addition to weathering of these walls due to ingress of flood water.

Figure 5. Inversion results of electrical resistivity tomography (ERT) profiles 1 to 4 across the area of the assumed extent of palace M750 (locations of profiles shown in Figures 1c and 2b): (a) inversion results for ERT profiles 1 to 4, carried out using Wenner arrays (electrode spacing 2 m); (b) inversion results for ERT profiles 1 to 4, carried out using dipole–dipole arrays (electrode spacing 2 m). This figure is available in colour online at wileyonlinelibrary.com/journal/arp.

Figure 5b shows that the dipole-dipole array has an increased resistivity resolution for mapping lateral changes. However, it is sensitive to noise and may therefore include them in the robust inversion. The inversion of dipole–dipole data produced anomalies with a more regular and sharper boundary than the anomalies of the Wenner array cross-sections. However, the high noise content was enhanced during the inversion, producing a resistivity section with poorer reliability. In general the dipole–dipole array produced shallower depths (less than 1 m) for most of the target wall anomalies compared with the depth (approximately 1.5–2 m) obtained by the Wenner array (Figure 5).

The dipole–dipole inversion result of ERT profile 5 is shown in Figure 6a. The vertically elongated rectangle-shaped anomalies with resistivity ranges 180–1500 Ω.m must be associated with the very hard dry sandstone wall foundations with depths between 0.3 m to 0.6 m. The interpreted walls are surrounded by material with relatively low resistivity (10–70 Ω.m), and underlain by very low resistivity material (1–20 Ω.m). The low values are associated with the variable humus, mud and clay soils covering the site.

Figure 6. Inversion results of high-resolution electrical resistivity tomography profiles: (a) profile 5 carried out using a dipole–dipole array (electrode spacing of 0.5 m), diagonally across the area of the assumed extent of palace M750 (locations are shown in Figure 2b); (b) profile 6 carried out using a dipole–dipole array (electrode spacing of 1 m), across the enclosure wall of the Royal City (Figure 3b). This figure is available in colour online at wileyonlinelibrary.com/journal/arp.

The inversion result of ERT profile 6 is shown in Figure 6b. The most prominent anomaly in this profile is displayed as a rectangular body with a horizontal length of about 5 m. It has a resistivity of about 2500 Ω.m, which is associated with the very hard, dry sandstone wall seen in the magnetic data. The inverted section indicates that the vertical extension of the enclosure wall is between 1 and 1.35 m, which correlates with the excavated depth of 1.5 m (Mohamed-Ali, 2007).

Discussion

The most interesting results in the current study were found in area-I southwest of palace M 750, which was a very important building located outside the confines of the Royal Enclosure. The dimensions and orientation of the detected building indicate that this building is likely to be the assumed southern part of palace M750 (Török, 1997). A test excavation (Figure 2c) was carried out in the area shown by the dashed box (Figure 2a). The excavation shows a small room with walls constructed from friable sandstone. Fired-brick remains were encountered in the entrance area of the excavated room where magnetic dipole anomalies were mapped.

The magnetic results encouraged us to carry out complementary electrical resistivity surveys to define both the vertical extension of the archaeological features and their degree of preservation in the wet geological setting. Details on these studies are found in Mohamed-Ali (2007).

The inversion results of ERT data show that the errors for the Wenner data are relatively less than those for the dipole–dipole data and therefore the results of the Wenner data are more reliable. This confirms the assertion by Drahor et al. (2008) that the Wenner configuration has a good signal-to-noise ratio for highly resistive topsoil conditions, but the relatively large spacing (2 m) used in some ERT profiles has a negative impact on the resolution of resistivity and depth values.

Robust (blocky) inversion produced anomalies of regular and sharp boundaries, particularly in dipole–dipole cross-sections, whereas the anomalies of the Wenner array cross-sections show a smooth and irregular geometry. The inverted Wenner and dipole–dipole ERT data show a low resistivity range (40–80 Ω.m) for the anomalies of sandstone walls. This probably is because the sandstone materials of these walls were damp and unconsolidated as a result of weathering. The other scattered irregular anomalies of the same range in the ERT sections may be attributed to the dry soil conditions. The highly exaggerated values of some anomalies may be associated with the dry sandstone walls of parts of the palace, which have been crossed diagonally. The 2500 Ω.m resistive anomaly in profile 6 probably is associated with the very hard, dry sandstone blocks of the royal enclosure wall. The measured values were a consequence of the contact resistance during the survey. We tried to overcome that problem by using salty water to inject current into the ground. We succeeded to some extent but still the effects are seen on the measured values. Moreover, these values might also be exaggerated by the numerical simulation of the inversion process, particularly as a result of the small spacing (0.5 m and 1 m) used for the dipole–dipole data collection. This array type and spacing focus the current in shallow depths and accordingly provided data that included the large resistivity contrast near the surface. Such a condition is a problematic in robust inversion using the Res2Dinv software. The current results correlate well with resistivity results under similar conditions revealed by Shaaban and Shaaban (2001) to depict the buried wall-like structures of hard limestone (12–6900 Ω.m) at the Tell El Rabi'a archaeological site.

The broader Wenner spacing produced relatively smooth anomalies. These results could not provide accurate estimation for the locations and dimensions of the interpreted walls seen in the magnetic image, but they can be used to highlight the applicability of ERT surveys for a rough assessment of the variation of the magnetic anomalies with depth.

The locations of the anomalies associated with the walls of palace M750 and their surroundings match to some degree the corresponding locations from the magnetic survey, suggesting that the dipole–dipole resistivity tomography data with 0.5 m spacing successfully defined the vertical boundaries of the wall structures. However, their thicknesses variations between 0.3 and 0.6 m shows poor correlation with that revealed (approximately 1 m) by excavation (Grzymski, 2005).

The very low resistivity values of both the sandstone walls and the surrounding materials may raise the question of the relationship of this building to the proposed Nile channels (Grzymski, 2005) and thereby support the classic references to the island of Meroe, defined as a large area encompassed by the rivers Nile, Atbara and Blue Nile (Figure 1a) as proposed by Bradley (1982).

Conclusions

The integration of magnetic gradient and electrical surveys for archaeological prospection in the central area of Meroe City successfully discovered unexcavated building structures and extended the area of known structures. Electrical resistivity tomography profiles selected on the basis of the magnetic results were used to provide a rough estimation for the locations and dimensions of the interpreted walls. The dipole–dipole survey results were more descriptive in determining the lateral extent of the subsurface features, but were not accurate owing to noise contamination and the large resistivity contrast near the ground surface. An ERT survey is recommended during or after rainy season when the damp soil condition is favourable to overcome the problem of poor injection of the electric current into the subsurface. Test excavation in Meroe confirmed the interpretation of the magnetic and resistivity anomalies of the extension for the palace M750, thereby proving the usefulness and applicability of these methods for archaeological prospection in Meroe and, accordingly, may encourage geophysical prospection of archaeological sites with similar conditions in Sudan.

Acknowledgements

The authors would like to thank the DAAD, Germany and the Islamic Development Bank, Jeddah, KSA for funding the project. Special thanks to Professor Badr el Din Khalil (Al Neelain University, Sudan), Professor Hans Burkhardt (TU, Berlin) and Professor He-Ping Sun (CAS, China) for help and cooperation during the course of this project. The Geological Research Authority of Sudan (GRAS) is much appreciated for providing the ERT system. Help from the German Archaeological Institute (Dr Simone Wolf and Dr Hans-Ulrich Onasch) and Dr Colin Sargent at Durham University is also greatly appreciated. Thanks are due to Dr Salah Mohamed Ahmed and Dr Abdelrahman Ali for encouragement and valuable comments and also to Professor Ali Osman and Dr Intisar El Zein (U of K, Sudan), and to the National Corporation of Antiquities and Museums (NCAM, Sudan) for providing access to the archaeological sites.