Geophysical prospection of a bronze foundry on the southern slope of the acropolis at athens Greece.

Archaeological Prospection
Archaeol. Prospect. 18, 27–41 (2011)
Published online in Wiley Online Library
(wileyonlinelibrary.com/journal/arp) DOI: 10.1002/arp.402
Geophysical Prospection of a Bronze Foundry
on the Southern Slope of the Acropolis at
Athens,Greece
MATTHIAS LEOPOLD1*, EVEY GANNAWAY1, JO«RG VO«LKEL1, FLORIAN HAAS2,
MICHAEL BECHT2,TOBIAS HECKMANN2, MARKUS WESTPHAL2 AND
GERHARD ZIMMER3
1
Technische Universitt Mˇnchen, Center of Life and Food Sciences Weihenstephan, Geomorphology
and Soil Science, Freising-Weihenstephan 85350, Germany
2
Katholische Universitt Eichsttt-Ingolstadt, Physical Geography, Eichsttt 85072, Germany
3
Katholische Universitt Eichsttt-Ingolstadt, Classical Archaeology, Eichsttt 85072, Germany
ABSTRACT
The sanctuary of the Acropolis of Athens in Greece provided one of the first monumental bronze statues some 2500
yearsago, whichwas dedicated to the goddess Athena.Duringrecent decades, important understanding ofthe statue’s
manufacturing processes has been achieved byarchaeological studies, and the former production site has been identified on the southern slope of the Acropolis.Two major bronze production pits have been detected and one was excavated in 2001 and 2006 and was found in an unexpected location. Therefore, in 2010 a geophysical survey of the
wholeproduction sitewascarried out for the first timeinorder to eitherrevealor to excludeany further sitesofthebronze
foundry complex. A combination of different geophysical methods was applied to survey the subsurface; magnetometry (MAG), two- and three-dimensional electrical resistivity tomography (ERT), as well as two- and three-dimensional
ground-penetrating radar (GPR).Two major anomalies have been identified in the processed data, which provide evidence foradditionalproduction sites.One was a known siteidentifiedin a test trenchin 2001, and our surveyhas outlined
the extent of the former pit. The other anomaly, which was detected by ERT and GPR, was 8^10 m in length and 2^
3 m in width and is oval-shaped and about 2.5 m deep. Steep vertical walls, together with two narrow points at the ends
ofthe pit, which could reflect formerentrances, were identified.Virtual ERTand GPR models generated from cross-sections ofa ground-based LiDAR scan of the 2001and 2006 excavated pit helped to interpret and understand the geophysical data of anomaly 2. This anomaly was finally interpreted as a newly detected production pit of the bronze
foundry complex, and based on these findings new excavations are planned. Copyright # 2011JohnWiley & Sons,Ltd.
Key words: magnetometry; electricalresistivity tomography; ground-penetrationradar; Acropolis; bronze foundry;
geophysical prospection
Introduction
The southern slope of the Acropolis of Athens with
the Sanctuaries of Dionysus and Asclepius was late
in becoming a subject of archaeological research. This
was due in large part to the accumulating debris from
the excavations of buildings on the Acropolis plateau
that was dumped in this area. It was not until 1877–78
* Correspondence to: M. Leopold, Technische Universität München,
Center of Life and Food Sciences Weihenstephan, Geomorphology
and Soil Science, Freising-Weihenstephan 85350, Germany.
E-mail: leopold@wzw.tum.de
Contract/grant sponsor: German Research Foundation DFG Az.
ZI335-7/2.
Copyright # 2011 John Wiley & Sons, Ltd.
that S. Koumanoudis, at the request of the Archaeological Society of Athens, excavated the area between
the Theatre of Herodes Atticus and the Sanctuary of
Asclepius down to the original ground level (Koumanoudis, 1878). In doing so he discovered a large pit
with remnants of bronze, bronze slag and fired clay
bricks (Zimmer, 1990). In 1963–64 Nikolaos Platon
implemented a cleaning of the area and identified the
heavily degraded pit as an establishment in which
the casting mould of a bronze statue was constructed,
dried, burnt out and filled with the liquid melt (Platon,
1964).
Based on new findings of a foundry for a cast bronze
statue in Rhodes (Kantzia and Zimmer, 1989) and a
new interpretation of the published drawings of Platon
Received 27 September 2010
Accepted 3 January 2011
28
(Zimmer, 1990), the 1st Ephoria of Prehistoric and
Classical Antiquities excavated a test trench to the
east of this pit in 2001. Instead of what was assumed
to be a wall, another casting pit was found with
well-preserved fixtures of clay bricks from which the
working process and applied technology can be
reconstructed (Zimmer, 2009). We now know that
portions of a cast were constructed in both pits for a
monumental statue of Athena, which was intended for
placement in the central area of the Acropolis in the
middle of the fifth century BC. Obviously connected to
this was the question of whether even more installations of the bronze foundry complex exist to the west
of the pit?
Geophysical prospection using the different methods
of magnetometry (MAG), electrical resistivity tomography (ERT) and ground-penetrating radar (GPR) was
therefore conducted in the adjacent area to the west in
the spring of 2010. In order to enhance the effectiveness
of an archaeological interpretation of the geophysical
results, we introduce an approach that allows us to
compare the results from the field data with the results
yielded from a virtual model. The model uses the shape
of the pit excavated in 2001 and 2006, and virtually
conducts ERT and GPR surveys using the same
configurations as in the field surveys. The example
helps archaeologists to understand the strengths and
weaknesses of geophysical prospection data.
Study site
The area studied is located on the southern slope of the
Acropolis of Athens in southern Greece at an altitude
of 120 m a.s.l. (Figure 1a–c). The Mediterranean climate
of Greece ensures typically warm, dry summers and
mild winters (Matyasovszky et al., 1995). The mean
annual temperature in Athens is 18 8C. The mean total
annual rainfall is 414.1 mm yr1, while the mean
annual rainfall depth is 390 mm (Koutsoyiannis and
Baloutsos, 2000; World Meteorological Organization,
2007). Athens is located in the southwest of the Athens
Basin, also known as the Attica Basin.
The geological bedrock of the Athens region is the
‘Athens Schist’ Series, which consists of schists,
sandstones, chert, shales, marls and limestones that
exhibit low-grade metamorphism on the local scale.
The hills scattered around the city, including the
Acropolis, are covered by the variably thick, compact
Tourkovounia Limestones known as ‘crest limestones’
that exhibit a thickly bedded to massive nature and are
Upper Cretaceous in age. In places the Attica Basin is
covered by Tertiary clay, marl, sandstone and
Copyright # 2011 John Wiley & Sons, Ltd.
M. Leopold et al.
conglomerate deposits, scree and talus cones, Quaternary alluvial deposits, and historic fill, known as the
archaeological layer of the city (Koukis and Sabatakakis, 2000; Marinos et al., 2001; Karfakis and Loupasakis,
2006).
The lithology of the Acropolis of Athens, and the
study site in particular, can be more specifically
described. Thick-bedded to massive grey Tourkovounia Limestones form the hill itself. The limestones are
up to 40 m thick and have been intensively faulted and
karstified. A reddish conglomerate exists along the
base of the hill, as well as large outcrops of a schist–
sandstone–marl series that constitute the upper layers
of the ‘Athens Schist’ Series. In some locations grey
limestones can be identified that are intercalated in the
schist-sandstone-marl series. An eluvial mantle covers
much of the slopes of the Acropolis of Athens, and
although it is widely extensive, its thickness is rarely
greater than 1 m. This mantle developed as a weathering product of the schist–sandstone–marl series
below (Andronopoulos and Koukis, 1976). A general
geological map of the southern slope of the Acropolis
is provided as Figure 1d (adapted from Andronopoulos and Koukis, 1976). The initially discovered pit was
constructed by removal of this eluvial mantle and
digging into the unweathered schist–sandstone–marl
series.
The Attica Basin, bounded by Mount Parnitha in the
north, Mount Aegaleo in the west, Mount Pentelli in
the northeast, and Mount Imittos in the east, lies in one
of the most tectonically active regions in Europe. The
greater region of the Aegean has experienced intense
active extension ever since the Upper Miocene, and
this neotectonic deformation is expressed in the basin
and range topography to which the Attica Basin
belongs (Lekkas, 2000; Tsodoulas et al., 2008; Krohe
et al., 2010). The extensional domain that exists in the
Attica Basin has resulted in widespread faulting,
predominantly expressed as normal faults, such as the
ENE–WSW and SE–NW striking normal faults that can
be traced around the Acropolis Hill. In many locations
these faults intersect the Tourkovounia Limestones
and have been subjected to intensive karstic weathering, resulting in the development of karstic voids
throughout the landscape. The voids are typically 0.5
to 2 m wide, but isolated voids have been found up to
15 m in width, and they can be partially or completely
filled by clays or stalagmitic material (Ganas et al.,
2005; Karfakis and Loupasakis, 2006). The normal
faulting and the karstic weathering present in the
study site region are important to acknowledge
because of the possibility that such features might
be mistaken for an anthropogenic or archaeological
Archaeol. Prospect. 18, 27–41 (2011)
DOI: 10.1002/arp
Geophysical prospection at Athens’ Acropolis, Greece
29
Figure 1. (a) Site maps ofthe study site, whichislocatedin Greeceinthe centre ofthe Cityof Athens (A). (b) Image ofthe Acropolis of Athensas seen
from the south.The two arrows point to the study site on the south slope of the Acropolis, northeast of the Roman Odean of the Herodes Atticus.
(c) Detailed plan of the study site with indication of all survey lines and areas and their location beside the previously excavated pits from 1963 and
2001. (d) Geological map of the larger area around the study site (modified after Andronopoulos and Koukis,1976). Location of the study site in the
northwest corner of the map is indicated by white arrows.This figure is available in colour online at wileyonlinelibrary.com/journal/arp
structure, such as a casting pit, when using geophysical prospection methods.
Methods
The geophysical methods of magnetometry, electrical
resistivity tomography and ground-penetrating radar
were employed along with sedimentological analysis.
Detailed parameters of each survey and the postprocessing methods applied are provided in Table 1.
Additionally, the parameters of the ground-based
LiDAR scan of the excavated pit are given.
Sedimentological analysis
Sedimentological analyses included texture classification by a combination of sedimentation and sieving
Copyright # 2011 John Wiley & Sons, Ltd.
for the clay and silt fraction and the sand fraction,
respectively; pH determination in CaCl2 with a glass
electrode from WTW (pH521); electrical conductivity
calculation by pH/Cond 340i from WTW; and
magnetic susceptibility measurements by the Kappabridge MS2 from Bartington at 0.47 MHz. Hartge and
Horn (1992) and Kretschmar (1996) describe in greater
detail the methods and procedures used in these
analyses.
Magnetometry
Magnetometry detects magnetic anomalies that have
been added to the natural pattern of Earth’s magnetic
field. These anomalies can be the result of a variety of
buried anthropogenic and archaeological features such
as ditches, walls, kilns, pits, etc. (Clark, 2001; Gaffney,
Archaeol. Prospect. 18, 27–41 (2011)
DOI: 10.1002/arp
Copyright # 2011 John Wiley & Sons, Ltd.
Total surveyed
Sampling
interval [m]
Total number
of samples /
traces
Acquisition
software
Processing
software
Processing
filter sequence
Manufacturer
ReflexW 5.5
X-flip; Subtract
mean dewow;
Correct maximum
phase;
Move start time;
Background removal;
Trace incr.-resampling;
Time cut;
fk migration;
Gain function;
Geotest 2.20 m
RES2DINV 3.55
Wenner a (2D);
Pole-Pole (3D);
Examine bad data points;
Topography;
Least-squares
inversion; converg.
limit 3 %; max. no.
of inversions: 5;
^
Geoplot 3.00
Mastergrid;
Composit;
Defect Removal: Clip;
Deslope;
Edgematch;
Despike;
Smoothing:
LP-filter, Interpolate;
GroundVision1.4.5
385 per grid 40000
1491
480
0.05
Model parameter:
Wenner Alpha;
No. of electrodes 25;
No. app. resistivity
levels:10;
No. resistivity
val.: 5; No. of grid
rows:14;
RES2DMOD 3.01
1652
^
1.0
^
2D-Line
Ramac CUII
20 20 m
& 10 8 m
1600
180
2.0 & 1.0
Loke (2002)
ERT Model
MAL— Geoscience
GPR
^
1.0
Lippman Geophysik.
Messgerte
4-Point light hp
2D-Lines (10) maximum 3D-Grid [m]
distance of curr. - pot.
12 12 &
Electrodes 8^10 m
66
ERT
480
0.5 0.25
Geoscan
Research
Model
FM 36 Gradiometer
Survey grid [m] 20 20 m &
10 8 m
MAG
Riscan Pro1.50
Raw Data/points
scanned: 4 906 121
LMS Z420i
12 m Resolution
of point cloud after
postprocessing:
870 000 points;
Resolution DEM:
0.05 m
30
0.02
Riegl
g.b. LIDAR
Riscan Pro/Laserdata
FD Model
LIS Desktop
parameter:
reflecting;
exploding
reflector;
kuepper; 250 MHz;
delta x: 0.05;
delta t: 0.1;
fk migration
ReflexW 5.5
400
^
0.05
^
2D-Line
Sandmeier (2007)
GPR Model
Table 1. Comparison of the different methods, equipment used, specific parameters of the field surveys, applied software and filter sequences of the multi-method survey
30
M. Leopold et al.
Archaeol. Prospect. 18, 27–41 (2011)
DOI: 10.1002/arp
Geophysical prospection at Athens’ Acropolis, Greece
31
2008). In the case of archaeological research, the
successful application of magnetometry has been
widely acknowledged and accepted (Witten, 2006;
Aspinall et al., 2008). Magnetometry was conducted in
two grids (20 20 m and 10 8 m) using a Fluxgate
Gradiometer FM36 from Geoscan Research in a zig-zag
mode. The sample interval was 0.25 m along the lines
and the lines were 0.5 m apart.
However, the results of the MAG are not presented
in this paper as the whole image was dominated by
many magnetic dipoles produced by numerous metal
objects, including electric cables, water pipes, flood
lights and fire hydrants. The high magnetic anomaly
values of more than 100 nT from the large metal objects
overwhelm adjacent weak anomalies that are characteristic of archaeological features.
Electrical resistivity tomography
Electrical resistivity methods are based on spatially
detecting subsurface electrical properties of the
earth. A measureable electric potential field develops
when current is introduced into the ground. Changes
in sedimentary structure, often associated with
archaeological features, correspond to changes in the
apparent electrical conductivity (e.g. Kampke, 1999;
Burger et al., 2006). While ERT is still less common in
archaeology than other electrical resistivity methods,
it has been more frequently applied over the past few
years (Papadopoulos et al., 2009; Tsokas et al. 2009;
Valois et al., 2010). Electrical resistivity tomography
was also used at the southern wall of the Acropolis by
Tsokas et al. (2006).
Using the 4-Point light hp induced polarization
earth resistivity meter from Lippmann, the apparent
resistivity was measured along two-dimensional
lines or within a three-dimensional grid, and these
values are later inverted to specific resistivity values
(Lippmann, 2008). Two-dimensional ERT lines ranged
in length from 20 to 34 m and were spaced 2 m
apart. Electrode spacing was 1 m along the lines and a
Wenner array was applied. Using a convergence limit
of 3% resulted in a maximum of five iterations with
RMS errors of 2.3–6.7%. Three-dimensional ERT was
conducted in two grids (12 12 m and 6 6 m).
Electrodes were spaced at 2 m apart in the 12 12 m
grid, and 1 m apart in the 6 6 m grid. A pole–pole
configuration was used.
Ground-penetrating radar
Ground-penetrating radar is an impulse reflection
method. In this survey, electromagnetic waves were
emitted into the ground at a frequency of 250 MHz.
Different dielectric properties of the subsurface
cause the waves to reflect or refract in a specific
pattern. Varying sedimentary structures as well as
anthropogenic and archaeological features can be
detected (see Leckebusch, 2003; Conyers, 2004; Leopold and Völkel, 2004). The radar signal is traced over
a certain period of time and transmitted at different
positions along the zig-zag survey, which allows for
the calculation of a three-dimensional dataset. Time
slices are then calculated from these three-dimensional
datasets, and matched to different depth profiles.
Ground-penetrating radar was conducted in two grids
(20 20 m and 10 8 m) using the Ramac CUII from
MALÅ Geoscience. The sample interval was 0.02 m
along the lines and the lines were 0.5 m apart. The
theoretical vertical resolution using 250 MHz antennae
is between 0.08 and 0.16 m at v ¼ 0.08 m ns1 according
to Sheriff and Geldhard (1982).
Virtual modelling
RES2DMOD version 3.01 from Loke (2002) was used
to perform forward modelling and to produce an
apparent resistivity pseudosection based on a virtual
model. The parameters of the model were derived
from field measurements and laboratory analysis
of the sediments as given in Table 2.
ReflexW version 5.5 from Sandmeier (2007) was
used to produce a radar image based on a model. The
vertical borders of the model were derived by x–
z profiles of a laser scan model of the casting
pit excavated in 2001 and 2006. Modelling parameters
were similar to those used during the field survey.
Ground-based light detection and ranging image
The extremely well-preserved pit, in combination
with the complexity of its internal assembly, led to
Table 2. Sedimentologicalanalysis; Acropolis1representstheinsitumaterialofsilt andclaymarls,Acropolis 2 representsthepit-fillmaterial
Sample
Acropolis1 (marl)
Acropolis 2 (pit filling)
% material
> 2 mm
Sand
Silt
Clay
pH
Munsell
colour
Conductivity
(mS m-1)
Magnetic
susceptibility
er
28.0
50.8
17.7
36.5
68.7
46.3
13.6
17.3
7.9
7.7
2.5 Y 7/3
10 YR 6/6
6.5
1.1
17.6
45.0
8.5
5.0
Copyright # 2011 John Wiley & Sons, Ltd.
Archaeol. Prospect. 18, 27–41 (2011)
DOI: 10.1002/arp
M. Leopold et al.
32
the decision to survey and document the shape of
the pit with the help of a ground-based LiDAR
scan. A second reason to survey the pit was the
fast weathering processes that take place on such a
terrestrial object.
To produce a detailed and accurate three-dimensional view (for details see Table 1), the former bronze
foundry (2001 and 2006 excavated pit) was surveyed in
2008 using a Terrestrial Laserscanner (Riegl LMS
Z420i). Because of a glass roof covering the former
bronze foundry, the scanning had to be done from
inside the pit. To reduce shadowing effects from the
platform in the middle of the pit and to obtain a
detailed and complete view, the pit had to be scanned
from seven single scan positions (see Figure 6).
After the fieldwork, these single scan positions were
matched together by post-processing using RiscanPro
software. The resulting point cloud (4 906 121 points)
was thinned to 870 000 points and then exported from
RiscanPro as an ASCII-file. The further processing was
done using the GIS package Laserdata LIS Desktop,
which includes the open source GIS SAGA. With
LIS Desktop and SAGA, a digital elevation model
(DEM, resolution 0.05 m) of the pit was produced, and
on the basis of this DEM, a grid of cross-sections (grid
space 0.5 m in the vertical and horizontal directions)
through the pit was derived. Cross section number
6 was chosen for the further analyses (see Figure 6).
Results and interpretation
Sediment parameters
The Athens marl that underlies the study site as the
basal geological sediment consists mainly of silt and
clay together with a varying content of carbonaceous
gravel (Andronopoulos and Koukis, 1976). Some basic
sediment parameters of the marl are given in Table 2.
Of the fraction < 2 mm, 68.8% is silt, whereas sand
accounts for 17.7% and clay 13.6%. The pH value of 7.9
is high due the carbonaceous setting. The colour of the
pale yellow marl is best represented by a Munsell
colour of 2.5 Y 7/3. Rather high conductivity values for
the marl were measured in the field and the laboratory
of 8.5–4.1 mSm1. Magnetic susceptibility reached
a value of 17.6 105 SI. The X-ray diffraction analysis
of the clay fraction of the phyllosilicates showed a
high percentage of primary chlorites together with
smectites, illites and kaolinites.
The preservation of the excavated pit from 2001 and
2006 also offered the possibility to obtain physical and
chemical data about the backfill. The texture of this
Copyright # 2011 John Wiley & Sons, Ltd.
anthropogenic sediment is much coarser, with more
than 50% in the coarse fraction (> 2 mm). The fine
fraction (< 2 mm) is calculated as 36.5% sand, 46.2%
silt and 17.3% clay in one sample, and also shows a
general coarser tendency as compared to the marl.
The pH is high with a value of 7.7 and the Munsell
colour is 10 YR 6/6. The conductivity is calculated as
1.8–0.6 mSm1, which is much lower than in the
underlying marl. In contrast, the magnetic susceptibility is higher than in the marl and reaches values
of 45.0 105 SI. The X-ray diffraction analysis on
the clay mineralogy showed essentially the same clay
mineral distribution in the backfill as in the marl
sample, which indicates that parts of the backfill
consist of the formerly excavated marl.
On the whole, the physical and chemical properties
of the marl and the backfill of the pit show large
contrasts, which is a necessary precondition to receive
positive results in any of the applied geophysical
prospection methods.
Electric resistivity tomography
Figure 2 shows the results of the different twodimensional lines along the study site. All lines have
the same survey parameters and inversions were
calculated using the same configurations given in
the methods sections. ERT-Line 1 best represents
the sedimentological conditions of the site. A distinct
zone of about 0.8 m thickness with resistivity values
of ca. 200 to 1000 Vm lies on top of a zone with
much lower resistivity ranging from 40 to 80 Vm.
The gradient between the two zones is sharp, which
most likely indicates a change in sedimentology. At
depths between 3 and 4 m a more gradual shift
from 50 to 80 Vm to slightly higher resistivity values
of up to 200 Vm can be observed. This general vertical
distribution of the subsoil conductivity is found
throughout ERT-Lines 1 to 7. However, ERT-Lines 3,
4 and partly 5 indicate a high resistivity zone between
500 and 900 Vm down to nearly 1.9 m basically along
the first 10 m of the lines.
ERT-Lines 8, 9 and 10 differ from the previously
described pattern. ERT-Line 8 shows a sharp zone of
high resistivity values between 8 to 16 m on the line.
Values of more than 1300 Vm can be observed down to
an approximate depth of 2 m. ERT-Line 9 displays a
similar resistivity pattern. Again, a high resistivity
anomaly can be identified between 12 and 23 m on the
line down to a depth of roughly 3 m. This anomaly is
characterized by value of 1000 to more than 1300 Vm
locally. The resistivity gradients below the anomaly
are sharp and they are indicated in the image by yellow
Archaeol. Prospect. 18, 27–41 (2011)
DOI: 10.1002/arp
Geophysical prospection at Athens’ Acropolis, Greece
33
this anomaly could not be displayed in Figure 2,
although it should be around 3.5–4 m. However, again
the left (southern) and the right (northern) side of the
anomaly are well displayed by the high resistivity
gradient between 8 and 18 m. Resistivities in the
described zone are up to 1100 Vm. In ERT-Line 10
there is another spatially more limited high resistivity
zone between 0 and 6 m on the line, which has a depth
of about 1.5 m and reaches values of up to 1300 Vm.
After the large anomaly had been identified in the
two-dimensional sections numbered 8, 9 and 10, a
three-dimensional ERT survey was conducted directly
above it (compare Figure 1c). The inversion resulted in
depth slices of the resistivity distribution as given in
Figure 3. Here, each layer represents a 50-cm-thick
volume. In Layers 2 and 3 and partly in 4, which cover
a depth range down to about 2.0 m, a pronounced
anomaly is visible in the east side of the survey grid
(right side of the depth slices). It shows a sharp edge to
the west where resistivity values drop from over
1000 Vm to less than 120 Vm within several decimetres. The anomaly has an elliptic shape and the
structure that causes it seems to be deeper in the
central part – information which is inferred from the
two-dimensional survey.
Ground penetrating radar
Figure 2. Visualization of the two-dimensional ERT lines measured at
the study site. For the locations of ERT-Lines 1 to 10, see Figure 1c.
ERT-lines1to10 are displayed fromtopto bottom.Fora bettercomparison, all figures use the same metric scale and contour values. This
figureisavailableincolouronlineat wileyonlinelibrary.com/journal/arp
colours, which are roughly 450 Vm. Below 3 m the
values drop to 250 Vm and less. ERT-Line 10 is the last
of the lines that shows this well developed resistivity
anomaly. Due to the shorter line length and the
accordingly smaller penetration depth, the bottom of
Copyright # 2011 John Wiley & Sons, Ltd.
Figure 4a represents one of the 40 radar lines that have
been measured on the site, and the filter sequence of
Table 1 was applied during post-processing. It is
situated at x ¼ 16.5 on the survey grid and has the
identical course as ERT-Line 9 (line a–a’ in Figure 4c).
The radar image is characterized by a high amplitude,
horizontal to subhorizontal continuous reflection
pattern over several metres down to approximately
50 ns of the two-way-traveltime (TWT). At about 30 ns
and between 0 to 10 m and 17 to 20 m along the line,
most of the radar signal is absorbed and below only
high frequency noise is visible. This general pattern is
disrupted between 11.5 and 16 m, where from 30 ns
down to about 52 ns continuous reflections are visible.
Another weak reflection, which is indicated by a
dotted line in Figure 4b, starts at 0 m and 50 ns, rises to
35 ns at 4.8 m and falls thereafter to 60 ns at 20 m.
Figure 4b highlights the general reflection pattern
choosing only the most prominent and continuous
high-amplitude patterns, and it further highlights a
clearly visible anomaly (anomaly A2) between 11.5
and 16 m.
The whole dataset of 40 lines has been interpolated
to a three-dimensional dataset that allows several time
slices to be displayed, which can be transferred to
Archaeol. Prospect. 18, 27–41 (2011)
DOI: 10.1002/arp
34
M. Leopold et al.
Figure 3. Visualization of depth slices of the three-dimensional ERT measured. For the location of ERT-Grid 1, see Figure 1c. Each depth slice
represents a 50-cm-thick volume of the survey area. Note the sharp and narrow resistivity gradient in the middle of the slices between 0.5 and
2.0 m depth.This figure is available in colour online at wileyonlinelibrary.com/journal/arp
depth slices. Figure 4c shows a time slice at 25.2 ns
(TWT), which corresponds to a depth slice of
0.85 m at a velocity of v ¼ 0.07 m ns1. Two anomalies,
which have been outlined by the high-amplitude
concentration in these areas, are visible. The locations
correspond with the anomalies detected in the
two- and three-dimensional ERT surveys.
Interpretation and validation using a virtual
modelling approach
We could not yield any data using MAG that would
allow for an archaeological interpretation at this
study site. However, the ERT survey as well as the
GPR survey yielded results that clearly allow for an
archaeological interpretation. The location of anomaly
A1, detected by both ERT and GPR, corresponds
Copyright # 2011 John Wiley & Sons, Ltd.
with the location of a former excavation, where an
archaeological test trench detected a shallow pit as
indicated in Figure 4c. The plan given by Koumanoudis in 1878 shows an L-shaped pit, which could not be
detected during the excavations by Platon in 1963. The
test trench in 2001 was conducted in order to clarify
this circumstance. However, the excavation only
showed findings of a typical casting pit, but the
outlines of the pit have been unknown until now. This
assures us that the high resistivity zone in the ERT
image and the high-amplitude continuous reflections
in the GPR image, both extending below 1 m depth,
correlate with an archaeological feature, a shallow pit
in this instance.
Anomaly A2 is different in shape and depth
compared with A1. It shows a more oval-like form
with a maximum length of 8–10 m and a maximum
Archaeol. Prospect. 18, 27–41 (2011)
DOI: 10.1002/arp
Geophysical prospection at Athens’ Acropolis, Greece
35
Figure 4. Graphicalresultsofthe GPR survey. (a) One of 40 two-dimensional GPRlinesafterpost-processing.Notethehighattenuationofthe signal
inthelowerpart oftheradarimage. (b) Themost prominent high-amplitudereflectionsof (a) fora better visualization. (c) Adepth sliceat 85 cm, which
was derived bya three-dimensionalinterpolation ofall 40 two-dimensionallines.Note the two anomalies (A1and A2) indicatedin the image.Narrow
points of the elliptic anomaly A2, indicated by the white arrows, possibly reflect entrances.
Copyright # 2011 John Wiley & Sons, Ltd.
Archaeol. Prospect. 18, 27–41 (2011)
DOI: 10.1002/arp
36
M. Leopold et al.
Figure 5. Three-dimensional model of the depth of the actual upper boundary of the marl as imaged by the GPR survey. Average depth of the marl
is between 0.7 and 0.8 m, but there are two anomalies (A1and A2), which have a greater depth. A1is a shallow pit excavated in 2001. A2 represents
an anomaly that is interpreted as a newly discovered casting pit. Narrow points of the elliptic anomaly A2, indicated by the white
arrows, possibly reflect entrances.The section a^a’ indicates the position of the GPR line shown in Figure 4.This figure is available in colour online
at wileyonlinelibrary.com/journal/arp
depth of around 2.5–3 m. Both methods generally
show the shape of A2. We interpret A2 as an
archaeological pit. This is also justified by the sharp
and unnatural looking border of the anomaly in the
three-dimensional ERT inversion image. Along linear
lines and within a few decimetres, the resistivity
values rise drastically, indicating different subsurface
materials.
The several two-dimensional radar lines have been
used to digitize the path of the boundary between
the clear continuous signals and the region where the
signal gets absorbed. This border is interpreted as the
contact of the weathered, anthropogenically altered
upper zone and the in situ marl with high clay and
silt content that absorbs the electromagnetic signal of
250 MHz within a few centimeters. Several x–y–z files
were chosen from the radar images to outline this
important boundary. The data were used to interpolate
the in situ surface of the unaltered marl, and the results
are introduced in Figure 5. The main depth of the marl
is between 0.7 and 0.8 m, but the two anomalies A1 and
A2 stand out in the image as having greater depth than
0.8 m (Figure 5). The excavation trench from 2001 and
2006 is visible at A1 as a shallow rectangular ditch
situated towards the north of the pit. The pit itself has
an irregular geometry. The findings point to bronze
casting production. Thus, the former use of the pit
may have been in conjunction with bronze casting,
where smaller forms were burned and filled with
metal. There are clues that for smaller forms, shallow,
sand-filled pits have been used (Zimmer, 1990).
A2 represents the detected pit, and the image in
Figures 4 and 5 clearly shows the oval shape with two
narrow points at the southeast and the northwest ends
of the pit, which could reflect the former entrances,
Copyright # 2011 John Wiley & Sons, Ltd.
similar to what was documented at the excavated pit in
2001 and 2006 (Figure 6).
In this pit a large wax moulded part was established,
covered with clay and surrounded with alternating
layers of loam. To perform these jobs, the two entrance
steps were necessary. To dry and heat the moulded
part, the interior size of the pit was decreased by the set
up of brick walls to convert the pit into a kind of
kiln. After the casting process, the outer layers and
the unfinished castings were removed. The bottom of
the moulded part, together with remnants of the shape
of the robe, were preserved in situ (Zimmer and
Hackländer, 1997).
However, no trace of any entrance stairs or other
details seen in the excavated pit is documented in the
ERT/GPR images or the three-dimensional reconstruction. Therefore, the questions arise as to whether
the pit in A2 is another typical example of a bronze cast
production pit like the others nearby, or if it is simply a
different kind. In general, the potential resolution of a
250 MHz antenna is high enough to portray details
such as a 30 cm wide and 20 cm high step from the
entrance stairs (compare methods section). Because we
could not carry out another survey with higher
data density and higher potential resolution, we
developed an approach that allows us to evaluate if
the yielded data could be from a production pit or
not. Therefore, we used the virtual model of the
pit produced by the laser scan and extracted several
x–z sections in different directions (Figure 6). We chose
one x–z section along the longitudinal axis of the pit as
a reference section that is valid for our modelling
approach. Several offsets, indicated by grey arrows,
are visible at the oblique northern and southern parts
of the line representing the surface of the excavated
Archaeol. Prospect. 18, 27–41 (2011)
DOI: 10.1002/arp
Geophysical prospection at Athens’ Acropolis, Greece
37
Figure 6. (Top) Three-dimensional view of an analytical hillshade of the 2001and 2006 excavated pit on the basis of the ground-based LiDAR data.
Theimage, whichwasproducedin 2008, includesthelocationsofthe scan positions. (Bottom) Cross-sectionnumber 6, ahorizontalsectionthrough
the excavated pit derived from the DEM, which was used for the later virtualmodel of the pit.Note: Greyarrows point to the scanned entrance steps.
entrance steps. Next we used this precise section of
the pit and virtually filled it with material using the
physical parameters from Table 2. These parameters
represent the conditions yielded during our field
survey combined with the results from the sedimentological analyses and the addition of some subsurface
noise, which we simply suggested by adding irregular
higher or lower electric resistivity values. These noise
values represent fallen rocks from the adjacent cliffs
or clay lenses from the underlying marl mixed in with
the subsurface. Finally, RES2DMOD version 3.01 was
used to model an apparent resistivity pseudosection
Copyright # 2011 John Wiley & Sons, Ltd.
and compare it with the resultant apparent resistivity
section achieved during the field survey (Figure 7). The
modelling parameters are given in Tables 1 and 2.
The modelled apparent resistivity section shows
similar values over large parts of the pseudosection
when compared with the field measurements. The
general shape of the apparent resistivity values in the
modelled pseudosection also shows a similar pattern
to the field-measured pseudosection.
ReflexW version 5.5 was used to produce a virtual
radar image at 250 MHz at a step rate of 0.05 m along
the same x–z section as the ERT model (model
Archaeol. Prospect. 18, 27–41 (2011)
DOI: 10.1002/arp
M. Leopold et al.
38
Figure 7. (a) The grid model and the specific values used in the RES2DMOD software (Loke, 2002). The outline of the pit (white line) equals the
section derived from the laser image in Figure 6. (b and c) The model was used to produce an apparent resistivity pseudosection as seen in (b),
and to compare it with the results of the apparent resistivity section (c), which was measured during the field survey.This figure is available in colour
online at wileyonlinelibrary.com/journal/arp
parameters compare Table 2). First, we used a
simple, three-layer model with a high contrast between
pit filling and underlying material based on the
parameters given in Table 1 and 2 (Figure 8a). Besides
remnants of diffraction tails, which remained even
after fk migration, the general outline of the pit
could be well portrayed by the configurations used.
Also the entrance steps are visible within the model
(see arrows in Figure 8b), as are most parts of the
complex geometry of the pit’s lower boundary. In a
second model, we added high background noise
by changing the parameters for the model in
Figure 8a. ‘Random layer’, which produces randomly
distributed heterogeneities on defined values, and
‘transition zone’, which extends and smoothes the
transition zone between the model layers, have been
applied. Figure 8c shows the outcomes of this model.
The radar image documents the general outline of the
pit by high-amplitude reflections within the pit
location, indicated by the white dotted line above a
Copyright # 2011 John Wiley & Sons, Ltd.
zone of relatively strong signal attenuation. However,
the radar image cannot display details such as the
entrance steps or the complex course of the pit bottom.
This discrepancy towards the model in Figure 8b is
primarily because random noise covers these details
and the signal attenuates with depth (Figure 8).
Discussion and conclusion
Our results clearly document the existence of an
anomaly (A2 in Figures 4c and 5), which is similar in
shape and depth to a pit that was detected in 2001 and
excavated in 2006 and to the older pit excavated in
1963–64. This was also corroborated by the outcomes
of the virtual models. The oval shape, together with
the sharp and vertically steep boundaries towards
the west and east, must be interpreted as a humanconstructed hole rather than a natural hole such as a
filled doline. While dolines, which are associated with
Archaeol. Prospect. 18, 27–41 (2011)
DOI: 10.1002/arp
Geophysical prospection at Athens’ Acropolis, Greece
39
Figure 8. (a) The three-layer model and applied parameters used for the virtual GPR survey based upon the same geometry as Figure 7a. (b) The
reflection pattern derived from a model with high contrasts and sharp boundaries between the several layers: black arrows point to the imaged
entrance steps. (c) Thereflection pattern derived from a modelwithlowercontrast and transitionalboundariesbetweenthe severallayersanda high
randombackgroundnoise.Zonesofhigh-amplitudereflectionsandzonesofhigh signalabsorptionareindicated.If backgroundnoiseishighenough,
details ofthe pit geometryare superimposed on otherreflections, but the generallocationofthe pit is stillvisible from deeperhigh-amplitude signals.
the natural karstic environment around the Acropolis
of Athens, must be discussed as a possibility for
causing the anomaly A2, they are likely to be rounded
to subrounded and have oblique, not vertical rims
(Šušteršič, 2006). As solution of limestone is the major
cause for a doline, this geomorphic form also can be
excluded from causing the anomaly A2 because marl
forms the geologic unit at the study area (compare
Figure 1). This clay-rich marl is unlikely to form
dolines by solution of the carbonate material.
We expected to identify possible entrance steps into
such a bronze manufacturing pit as were documented
during the excavations in 1963–64 and 2001, but
such features could not be detected in either the GPR
or the ERT surveys. However, as the virtual model
results showed, if there is high enough random noise,
such details cannot be detected by the methods and
configurations we used during our field survey. In
this situation, one would simply overestimate the
potentials of these methods.
Nevertheless, there is clear evidence that anomaly
A2 is another casting pit used to manufacture part of
the former statue of Athena. The shape and depth are
similar to the known casting pits, which lay just beside
Copyright # 2011 John Wiley & Sons, Ltd.
A2. The three-dimensional ERT image clearly documents a sharp vertical boundary towards the west side
of A2, which is similar to the archaeological results of
former excavations where the walls of the casting pits
were nearly vertical and have additionally been built
with brick (Zimmer and Hackländer, 1997). The two
narrow points at the southeast and northwest ends of
the pit most likely represent former entrances with
possible small narrow steps down to the bottom of the
pit.
The use of different geophysical methods, as shown
previously in other studies (e.g. Leopold et al., 2010;
Lowe and Fogel, 2010; Maio et al., 2010), was especially
valuable in this case because magnetometry failed, but
GPR and two- and three-dimensional ERT surveys
produced good results, which substituted for the other
method.
The newly discovered pit (anomaly A2) is of major
importance for our understanding of the bronze
casting workshop at this internationally important
site. We know that the statue was 30 ft or 9 m in height.
If one subtracts the head and the helmet, this results in
a length of 7.5 m for the body of the goddess statue.
Because the workers in antiquity could not heat that
Archaeol. Prospect. 18, 27–41 (2011)
DOI: 10.1002/arp
40
deep a pit, and particularly because this was the first
monumental statue built out of bronze, they had to
perform the casting process in several parts. Based on
the new finding, it seems logical that they decided to
produce three individual parts 2.5 m high and to place
each part in its own separate pit for the casting process.
The statue was assembled on the podium west of the
1963–64 excavated pit. Smaller parts were cast in the Lshaped pit (anomaly A1). Taking all of this into
consideration, we are now better informed about the
technology and the workflow patterns of this bronze
casting workshop. By excavating the newly discovered
pit we hope to find parts of the clay and loam casting
moulds, which would give us more clues of the visual
nature of the great Athena statue. This would be very
valuable because in the fourth century AD the statue
was taken to Constantinople and around AD 1200 it was
destroyed without an image ever made of it (Lundgreen, 1997).
Acknowledgements
We thank Dr Alexandros Mantis, head of the 1st
Emporia of Prehistoric and Classical Antiquities, for
permission to conduct the geophysical survey. Sophia
Moschonisisotis, the supervisor responsible, helped
with the pre-arrangements of the survey for which
we are very thankful. We further thank Efi Kasapoglou
who excavated the pit in 2001 and 2006 and supervizes
the restoration. The manuscript was improved by the
constructive comments of two anonymous reviewers
for which we are grateful. We also thank C. Gaffney for
editing the paper. The project was funded by the
German Research Foundation (DFG Az. ZI335-7/2),
for which we are grateful.
References
Andronopoulos B, Koukis G. 1976. Geologic Map of the
Acropolis Area, Athens. Institute of Geological and
Mining Research.
Aspinall A, Gaffney C, Schmidt A. 2008. Magnetometry for
Archaeology. AltaMira Press: Walnut Creek, CA.
Burger HR, Sheehan AF, Craig JH. 2006. Introduction to
Applied Geophysics: Exploring the Shallow Subsurface.
Norton & Company: New York.
Clark A. 2001. Seeing Beneath the Soil – Prospecting Methods
in Archaeology. Routledge: New York.
Conyers LB. 2004. Ground-Penetrating Radar for Archaeology. Altamira Press: Walnut Creek, CA.
Gaffney C. 2008. Detecting trends in the prediction of
the buried past: a review of geophysical techniques
in archaeology. Archaeometry 50(2): 313–336. DOI:
10.1111/j.1475-4754.2008.00388.x.
Copyright # 2011 John Wiley & Sons, Ltd.
M. Leopold et al.
Ganas A, Pavlides S, Karastathis V. 2005. DEM-based
morphometry of range-front escarpments in Attica,
central Greece, and its relation to fault slip rates.
Geomorphology 65: 301–319. DOI: 10.1016/j.geomorph.
2004.09.006.
Hartge KH, Horn R. 1992. Die physikalische Untersuchung
von Böden, 3rd edn. Spektrum: Stuttgart.
Kantzia Ch, Zimmer G. 1989. Rhodische Kolosse. Archäologischer Anzeiger 1989: 497–523.
Kampke A. 1999. Focused imaging of electrical resistivity
data in archaeological prospecting. Journal of Applied
Geophysics 41(2–3): 215–227. DOI: 10.1016/S09269851(98)00043-3.
Karfakis J, Loupasakis C. 2006. Geotechnical characteristics of the formation of ‘Tourkovounia’ Limestones
and their influence on urban construction – City of
Athens, Greece. IAEG2006 Paper number 794.
Koukis G, Sabatakakis N. 2000. Engineering geological
environment of Athens, Greece. Bulletin of Engineering
Geology and the Environment 59: 127–135. DOI: 10.1007/
s100640000058.
Koumanoudis S. 1878. Working report. Praktika.
Koutsoyiannis D, Baloutsos G. 2000. Analysis of a long
record of annual maximum rainfall in Athens, Greece,
and design rainfall inferences. Natural Hazards 29: 29–
48. DOI: 10.1023/A: 1008001312219.
Kretschmar R. 1996. Kulturtechnisch-bodenkundliches
Praktikum, Ausgewählte Labor- und Feldmethoden.
Institut für Wasserwirtschaft und Landschaftsökologie,
Cristian-Albrechts-Universität: Kiel.
Krohe A, Mposkos E, Diamantopoulos A, Kaouras G.
2010. Formation of basins and mountain ranges in
Attica (Greece): The role of Miocene to recent low-angle
normal detachment faults. Earth-Science Reviews 98: 81–
104. DOI: 10.1016/j.earscirev.2009.10.005.
Leckebusch J. 2003. Ground-penetrating radar: a modern
three-dimensional prospection method. Archaeological
Prospection 10: 213–240. DOI: 10.1002/arp.211.
Lekkas E. 2000. The Athens earthquake (7 September
1999): intensity distribution and controlling factors.
Engineering Geology 59: 297–311. DOI: 10.1016/S00137952(00)00119-8.
Leopold M, Völkel J. 2004. Neolithic flint mines in Arnhofen, southern Germany: a ground-penetrating radar
survey. Archaeological Prospection 11: 57–64. DOI:
10.1002/arp.222.
Leopold M, Plöckl T, Forstenaicher G, Völkel J. 2010.
Integrating pedological and geophysical methods to
enhance the effectiveness of archaeological prospection: the example of a Roman villa rustica near Regensburg, Germany. Archaeological Science 37(7): 1731–1741.
DOI: 10.1016/j.jas.2010.01.033.
Lippmann E. 2008. 4 Point light 10W, Technical Data
Operating Instructions Version 4.19. Available from:
< http://www.l-gm.de >.
Loke MH. 2002. RES2DMOD ver. 3.01, Rapid 2D Resistivity
Forward Modeling Using the Finite-Difference and
Finite-Element Methods. Available from: http://www.
geoelectrical.com .
Lowe KM, Fogel AS. 2010. Understanding Northeastern
Plains Village sites through archaeological geophysics.
Archaeological Prospection, 17: 247–257. DOI: 10.1002/
arp.394.
Archaeol. Prospect. 18, 27–41 (2011)
DOI: 10.1002/arp
Geophysical prospection at Athens’ Acropolis, Greece
Lundgreen B. 1997. A methodological enquiry: the Great
Bronze Athena by Pheidias. Journal of Hellenic Studies
117: 190–197.
Maio RD, Fedi M, Manna ML, Grimaldi M, Pappalardo U.
2010. The contribution of geophysical prospecting in
the reconstruction of the buried ancient environments
of the house of Marcus Fabius Rufus (Pompeii, Italy).
Archaeological Prospection, 17: 259–269. DOI: 10.1002/
arp.395.
Matyasovszky I, Bogardi I, Ganoulis J. 1995. Impact of
global climate change on temperature and precipitation
in Greece. Applied Mathematics and Computation 71: 119–
150. DOI: 10.1016/0096-3003(94)00148-W.
Marinos P, Bouckovalas G, Tsiambaos G, Sabatakakis N,
Antoniou A. 2001. Ground zoning against seismic
hazard in Athens, Greece. Engineering Geology 62:
343–356. DOI: 10.1016/S0013-7952(01)00035-7.
Platon N. 1964. Working report. Archaiologikon Deltion
19: Chron.1.
Sandmeier KJ. 2007. ReflexW ver. 5.5. Available from:
< http://www.sandmeier-geo.de/Reflex/reflexw.htm >.
Sheriff RE, Geldart LP. 1982. Exploration Seismology,
Volume 1: History, Theory and Data Acquisition. Cambridge University Press: Cambridge.
Šušteršič F. 2006. A power function model for the basic
geometry of solution dolines: considerations from the
classical karst of south-central Slovenia. Earth Surface
Processes and Landforms 31: 293–302. DOI: 10.1002/
esp.1244.
Papadopoulos N, Tsokas G, Dabas M, Yi M, Kim J,
Tsourlos P. 2009. Three-dimensional inversion of automatic resistivity profiling data. Archaeological Prospection 16(4): 267–278. DOI: 10.1002/arp.361.
Copyright # 2011 John Wiley & Sons, Ltd.
41
Tsodoulas IM, Koukouvelas IK, Pavlides S., 2008.
Tectonic geomorphology of the easternmost extension
of the Gulf of Corinth (Beotia, Central Greece). Tectonophysics 453: 211–232. DOI: 10.1016/j.tecto.2007.06.015
Tsokas G, Tsourlos P, Papadopoulos N, Manidaki V,
Ioannidou M, Sarris A. 2006. Non destructive ERT
survey at the south wall of the Akropolis of Athens.
Extended Abstracts of Near Surface 2006, Helsinki, Finland; B040.
Tsokas N, Tsourlos PI, Stampolidis A, Katsonopoulus
D, Soter S. 2009. Tracing Major Roman Road in the
Area of Ancient Helike by Resistivity Tomography.
Archaeological Prospection 16: 251–266. DOI: 10.1002/
arp.355.
Valois R, Bermejo L, Guérin R, Hinguant S, Pigeaud R,
Rodet J. 2010. Karstic morphologies identified with
geophysics around Saulges caves (Mayenne, France).
Archaeological Prospection 17: 151–160. DOI: 10.1002/
arp.385.
Witten AJ. 2006. Handbook of Geophysics and Archaeology.
Equinox: London.
World Meteorological Organization. 2007. Hellenic
National Meteorological Service – Athens. Available from:
< http://worldweather.wmo.int/063/c00177.htm/ >.
Accessed July 2010.
Zimmer G. 1990. Griechische Bronzegusswerkstätten. Zur
Technologieentwicklung eines antiken Kunsthandwerkes: Mainz; 226 pp.
Zimmer G. 2009. Ergasthria xalkoplastikh sthn
arxaı́a Ella
da (copper workshops in ancient Greece).
Egnatia 13: 211–223.
Zimmer G, Hackländer N. 1997. Der Betende Knabe –
Original und Experiment. Peter Lang: Fankfurt/Main.
Archaeol. Prospect. 18, 27–41 (2011)
DOI: 10.1002/arp