Space-Based Ground Penetrating Imaging

of the Ancient Labyrinth at Hawara

Mark Carlotto (markcarlotto.com)

Abstract
Space-based ground penetrating radar imaging of an enormous underground structure at
Hawara that could be the below-ground portion of the Labyrinth first mentioned by Herodotus
in the fifth century BCE is presented. Initial detection results from Sentinel-1 C-band synthetic
aperture radar (SAR) later confirmed and enhanced with L-band ALOS/PALSAR reveal the
presence of an underground structure roughly the same size as that estimated by Finders
Petrie. Comparing radar and optical images reveals uncorrelated SAR returns on both sides of
the Abdul Wahbi canal that could be the remains of an ancient settlement built from and over
the Labyrinth that was uncovered when the canal was built as well as the remains of the
Labyrinth itself.

1. Introduc+on
First mentioned by Herodotus in the fifth century BCE,1 the existence of an enormous
underground structure in Hawara thought to be the remains of an ancient temple known as the
Labyrinth, has been a subject of considerable speculation but only limited scientific research.
Confirming the existence of such a structure could provide new insight into the early history of
the Egyptian civilization. Previous research efforts have detected evidence of below-ground
features that appear to be walls or even chambers, however, the overall size and shape of what
might be underground remains a mystery.

Space-based synthetic aperture radar (SAR) operating at C- and L-band wavelengths are capable
of penetrating desert sands and can, under certain conditions, detect subsurface features not
visible to optical sensors. Early experiments to model and better understand the physics of
ground penetrating SAR were conducted in hyperarid regions such as the Mohave desert2 and
the Sahara3 using Seasat SAR and Shuttle Imaging Radar (SIR-A) that operated in the L-band.4

1
Herodotus, History, 2.148.
2
R. G. Blom, R. J. Crippen, and C. Elachi, "Detec@on of subsurface features in Seasat radar images of Means Valley,
Mojave Desert California," Geology (1984) 12 (6): 346–349.
3
Charles Elachi, Ladislav E. Roth, and Gerald G. Schaber, “Spaceborne Radar Subsurface Imaging in Hyperarid
Regions,” IEEE Transac3ons on Geoscience and Remote Sensing, Volume: GE-22, Issue: 4, July 1984.
4
The nominal frequency range of L-band is from 8 to 4 GHz (3.75 to 7.5 cm).
More recently, ALOS/PALSAR, another L-band SAR, was used to detect and characterize a
known manmade feature buried under sand deposits in Egypt’s Western Desert.5

In this paper we present results from two polarimetric SAR sensors: Japan’s ALOS/PALSAR and
the Sentinel-1 C-band SAR developed by the European Space Agency under the Copernicus
initiative that provide the first verifiable space-based evidence of an enormous underground
structure at Hawara. Section 2 provides a brief historical overview including a review of
previous scientific investigations of the Labyrinth. Our initial Sentinel-1 image findings are
presented in Section 3. Follow-up work that involved registering ALOS/PALSAR and Google
Earth color images to the Sentinel-1 SAR, fusing the SAR images to visually enhance the original
results, and using a color (optical) image to estimate the uncorrelated (below-ground)
component in one of the SAR images are described in Sections 4 and 5. Section 6 discusses our
findings and areas for future work.

2. The Labyrinth
The legendary Labyrinth of Hawara was brought to the attention of the Western world by
Herodotus in the fifth century BC. He describes an above-ground structure that he saw, and one
below-ground that he was denied access to by the Egyptians:6

Moreover, they decided to preserve the memory of their names by a common

memorial, and so they made a labyrinth a li[le way beyond lake Moeris and
near the place called the City of Crocodiles. I have seen it myself, and indeed
words cannot describe it; if one were to collect the walls and evidence of
other efforts of the Greeks, the sum would not amount to the labor and cost
of this labyrinth. And yet the temple at Ephesus and the one on Samos are
noteworthy. Though the pyramids beggar descrip^on and each one of them is
a match for many great monuments built by Greeks, this maze surpasses even
the pyramids. It has twelve roofed courts with doors facing each other: six
face north and six south, in two con^nuous lines, all within one outer wall.
There are also double sets of chambers, three thousand altogether, fi`een
hundred above and the same number under ground. We ourselves viewed
those that are above ground, and speak of what we have seen, but we learned
through conversa^on about the underground chambers; the Egyp^an
caretakers would by no means show them, as they were, they said, the burial
vaults of the kings who first built this labyrinth, and of the sacred crocodiles.
Thus we can only speak from hearsay of the lower chambers; the upper we
saw for ourselves, and they are crea^ons greater than human. The exits of the
chambers and the mazy passages hither and thither through the courts were
an unending marvel to us as we passed from court to apartment and from

5
Ahmed Gaber, Magaly Koch, M. Helmi Griesh, Motoyuki Sato, and Farouk El-Baz, “Near-surface imaging of a
buried founda@on in the Western Desert, Egypt, using space-borne and ground penetra@ng radar,” Journal of
Archaeological Science, Volume 40, Issue 4, 2013.
6
Herodotus, History, 2.148.
 apartment to colonnade, from colonnades again to more chambers and then
into yet more courts. Over all this is a roof, made of stone like the walls, and
the walls are covered with cut figures, and every court is set around with
pillars of white stone very precisely fi[ed together. Near the corner where the
labyrinth ends stands a pyramid two hundred and forty feet high, on which
great figures are cut. A passage to this has been made underground.
Sir Flinders Petrie’s 19th-century survey provides a rough idea of the scale of the above-ground
structure (Figure 1). His assessment was that the Labyrinth was maintained up to the second
century BCE – "at least as late as Kleopatra I, 193 B.C., the Labyrinth was still in royal care, and
probably being restored in some degree. Soon after that, ruin fell upon it, and in Pliny's time it
was ‘marvellously ravaged.’”7 Petrie’s reconstruction of the western half of the Labyrinth
(Figure 2) indicates a north-south length of approximately 526 feet. Squaring that (assuming the
structure is roughly square) yields an area of about 277,000 square feet.

In 2008, the Mataha expedition reported the discovery of below-ground features resembling
archaeological structures south of the Hawara pyramid of Amenemhet III:8

These features covering an underground area of several hectares, have the

prominent signature of ver^cal walls on the geophysical results. The ver^cal
walls with an average thickness of several meters, are connected to shape
nearly closed rooms, which are interpreted to be huge in number.
Very low frequency electromagnetic (VLF-EM) surveys revealed “spatially distributed elongated
and square shaped subsurface anomalies, which may identify the walls and rooms of the
labyrinth mortuary temple complex.”9 Figure 3 displays an approximate geolocation of an
electrical conductivity map computed from VLF-EM surveys in Google Earth.

The presence of possible archeological features below ground suggests the possibility that what
Petrie thought was the foundation of a large above-ground structure could instead be the roof
of an undiscovered structure below ground. As the VLF-EM and other surveys were conducted
over a limited area, it was not possible to determine the size and shape of the underground
structure.

Inspired by the discoveries of the Mataha expedition, a remarkable 3D reconstruction of the

site using proprietary space-based technology revealed two levels of subterranean chambers
over an extended area surrounding the pyramid of Amenemhat III (Figure 4). But unlike the
Mataha findings, no accompanying scientific paper describing either the methodology or the
data used was ever published making it difficult to corroborate their findings.

7
Flinders Petrie, The Labyrinth Gerzeh and Mazghuneh (1912). hcps://archive.org/details/ERA21
8
The Mataha Expedi3on (2008) hcps://issuu.com/yago1/docs/labyrinth_of_egypt_com__hawra_2015
9
Mohamed Ahmed Khalil, Abbas Mohamed Abbas, Fernando A. Monteiro Santos, Hany S.A. Mesbah and Usama
Massoud, “VLF-EM study for archaeological inves@ga@on of the labyrinth mortuary temple complex at Hawara
area, Egypt,” Near Surface Geophysics, 2010, 8, 203-212. doi:10.3997/1873-0604.2010004
3. Sen+nel-1 C-Band Detec+on Results
Figure 5 shows a C-band Sentinel-1 SAR image collected 7/6/2023 over a region in Western
Egypt containing the Faiyum Oasis.10 Hawara is at the extreme top right edge of the image.
Sentinel-1 transmits vertically polarized waves and receives horizontally and vertically polarized
waves that are processed to form VH and VV images. VV images tend to pick up height/vertical
features whereas VH images tend to be more sensitive to surface textures. Longer wavelength
L-band SAR can penetrate 2–5 m below the surface, yet observations in the L-band have been
reported to provide less detail than shorter wavelength C-band11 despite its shorter penetration
depth (0.5 m).12

Figure 6 overlays a section of the Sentinel image centered on the pyramid of Amenemhet III in
Hawara in Google Earth. Figure 7 displays the same area without the Sentinel image overlay.
Although there are a number of strong SAR returns in its vicinity, the lack of a strong response
from the pyramid itself is likely due to its highly eroded state. Several bright returns (A) north of
the pyramid are from high-voltage transmission line towers. Below the pyramid in what is
believed to be a mortuary complex are at least three returns that are rectangular in shape. One
of the delineated regions (B) is about 275,000 sq. feet – almost the estimated size as the
rectangular area excavated by Petrie. Another region (C) may be the continuation of the above
structure. A second rectangular area (D) west of the Abdul Wahbi canal is also evident in the
SAR image. The lack of visible structures in Google Earth imagery over these areas suggests the
possibility that these returns could be subsurface features.

4. Cross-Sensor Correla+on Analysis

Following up on the Sentinel-1 findings, ALOS/PALSAR was obtained over the same area along
with a visible image from Google Earth. Data were co-registered using tie points and resampled
to produce a stack of seven image layers:

1,2 – VV and VH Sentinel-1 C-band polarization images collected on June 29, 202313
3,4 – HH and HV ALOS/PALSAR L-band polarization images collected on August 17, 200714
5,6,7 – The three (RGB) bands of a visible image collected on June 2, 2013, downloaded from
Google Earth

With reference to Figure 9, where structures are evident in the C-Band VV-VH false color
composite (a), the actual VV-VH values are not significantly different from the background. The
same can be said for the L-band HH-HV returns (b). We can exploit correlations between the
two images by combining the Sentinel-1 VV and VH and ALOS/PALSAR HH and HV polarizations

10
hcps://en.wikipedia.org/wiki/Faiyum_Oasis
11
The nominal frequency range of C-band is from 1 to 2 GHz (30 to 15 cm).
12
Eman Ghoneim, Michael Benedej, and Farouk El-Baz, “An integrated remote sensing and GIS analysis of the
Kufrah Paleoriver, Eastern Sahara,” Geomorphology 139–140 (2012) 242–257.
13
S1A_IW_GRDH_1SDV_20230629T160447_20230629T160516_049203_05EAA4_EC85.SAFE
14
AL1_OESA_FBD_GDH_1P_20070817T202737_20070817T202754_008326_0611_0560_0000
using principal components analysis (Table 1). By doing so, the Labyrinth stands out from the
background in an image constructed from the first three principal components (c).

5. Nonlinear Background Es+ma+on

Since there do not appear to be indications of structures in the visible image, we can use it to
test if the SAR is detecting a subsurface feature by attempting to estimate the SAR from the
visible image. Specifically, we use the three RGB bands of the visible image to estimate the
Sentinel-1 VV polarization component:

!"($, &) = ) [+($, &), ,($, &), -($, &)]

where ) is a nonlinear function that minimizes the mean square estimation error:15

/ [!"($, &) − !($, &)]!

It can be shown that the optimal estimate is given by the conditional expected value:

∑ ! 2(!|+, ,, -)
!" =
2(!|+, ,, -)

which is illustrated graphically in Figure 10. By definition, the error image

4{$, &) = !($, &) − !"($, &)

is uncorrelated with the visible image. As shown in Figure 11, the greatest concentration of
positive errors (where the actual SAR backscatter exceeds the estimated backscatter) occurs in
areas B and D marked in Figure 8. That these errors are uncorrelated with the visible image
implies the returns in this part of the SAR image are from something that is not present in the
visible image – something that could be underground.

6. Discussion
We can go a step further to eliminate the possibility that the uncorrelated returns are from
surface relief features. Since the VV polarization responds to vertical features, areas that face
the radar would be brighter than those that face away. Two transects are shown in Figure 12.
The terrain slopes downward from north to south and east to west. The canal is too narrow to
be resolved by the Google Earth elevation data. However, other than from the canal’s eastern
side, there do not appear any other terrain features capable of generating radar returns in the
area of the Labyrinth.

15
Mark J. Carloco "Nonlinear background es@ma@on and change detec@on for wide-area search," Op3cal
Engineering 39(5), (1 May 2000). hcps://doi.org/10.1117/1.602496.
The downslope west of the canal is an interesting aspect of the terrain in this part of the Faiyum
that is extremely relevant to the present discussion. The original waterway, the Bahr Yusef,
connects Lake Moeris with the Nile River. In the past when rainfall was more plentiful, the lake
was much larger in area and the water level was much higher. As the climate in this part of the
world changed and the lake began to dry up, water was diverted from the Nile by way of the
Bahr Yusef. In order to supply water to the northeastern part of the Faiyum, which is higher in
elevation, secondary canals were constructed. A canal known as the Bahr Wardan located west
of Hawara flowed to the northeast. Eventually, because it had become filled with silt and was
no longer usable, sometime between the 6th and 13th century CE a new canal, the Bahr
Sharqiyyah, which is now known as the Abdul Wahbi, was evidently cut through the Labyrinth.
According to Bryan Kraemer:16

The evidence concerning the course of the medieval and modern canal, dug
directly across Hawara, and the posited loca^on of the ancient canal, going
around Hawara on the west side, suggests that a significant effort was made
some^me before the 13th cent. AD to move the canal. The Bahr Sharqiyyah
was cut through the archaeological site of Hawara with a width of
approximately 30 meters and a depth of approximately 13 meters for
approximately 1 kilometer. The task of excava^ng this canal would have
involved the removal of mud-brick remains, compacted stone debris from the
remains of the Labyrinth, and a significant amount of limestone bedrock.
He goes on to state that “archaeological surveys at Hawara have established that ceramic
debris on the surface of the southwestern part of the site indicates the presence of a
settlement here from the 3rd cent. BC until the 1st cent. AD. The modern canal [Abdul Wahbi]
cuts through the foundations of mud-brick structures of this settlement, and stone debris from
the 12th dynasty mortuary temple of Amenemhat III on which they were built.”

Radar returns that are clearly evident in the Sentinel-1 VV polarization have been shown to be
uncorrelated with visible surface features. It is therefore likely that they originate from a
structure that is below ground. That the returns are in the same area described above by
Kraemer suggests that they are probably from the legendary Labyrinth.

One must wonder, why no attempt has been made to reach the Labyrinth from the side,
starting at the level of the canal, which is well below ground level, and working outward (Figure
13).

Areas for future work include identifying and analyzing data sets over this area to better
understand the size, shape, internal details, and state of the underground structure, in
particular, to determine if, like the pyramid of Amenemhat III and mortuary complex to the
north, it is flooded and could therefore be in a state of decay. Other techniques such as radar

16
Bryan Kraemer, “The Meandering Iden@ty of a Fayum Canal: The Henet of Moeris / Dioryx Kleonos / Bahr
Wardan / Abdul Wahbi,” Proceedings of the Twenty-FiIh Interna3onal Congress of Papyrology, Ann Arbor 2007
American Studies in Papyrology (Ann Arbor 2010) 365–376.
Doppler tomography may be able to reveal hidden details as in the internal structure of the
Great Pyramid of Giza.17

Table 1 Sen*nel-1/ALOS/PALSAR principal components analysis where SVV and SVH are the Sen*nel-1 VV
and VH polariza*ons and AHH and AHV are the ALOS/PALSAR HH and HV polariza*ons.

Eigenvectors SVV SVH AHH AHV Eigenvalue

PC01 0.397538 0.366493 0.568364 0.620169 9545.133789
PC02 -0.664943 -0.515959 0.33982 0.419714 1992.201782
PC03 -0.302049 0.407889 -0.658312 0.555894 1218.024536
PC04 -0.555501 0.6581 0.357925 -0.36085 1094.55127

17
Filippo Biondi and Corrado Malanga, “Synthe@c Aperture Radar Doppler Tomography Reveals Details of
Undiscovered High-Resolu@on Internal Structure of the Great Pyramid of Giza,” hcps://arxiv.org/abs/2208.00811
Figure 1 Petrie’s survey of the area south of Amenemhat III (Plate xxxii).
Figure 2 Petrie’s plan of the remains of the above-ground por*on of the Labyrinth at the same scale as
Plate xxxii.
Figure 3 Map derived from VLF-EM data registered to Google Earth. Red colors are conduc*ve zones
(archaeological remains) and colors are resis*ve zones (host soil).
 Figure 4 3D reconstruc*on registered to Google Earth. Imagery derived from video.18

18
hcps://www.youtube.com/watch?v=shoWLpGmVJM&t=1109s
Figure 5 Sen*nel 1 C-band SAR image over Western Egypt. Red represents the VH polariza*on, green
represents the VV polariza*on and blue is an average of the absolute values of the two polariza*ons.
Figure 6 Por*on of Sen*nel-1 SAR image over Hawara. Red-green composite shows the VH polariza*on in
red and the VV polariza*on in green.
Figure 7 Google Earth background image over the same area as in the previous figure. The pyramid of
Amenemhet III is near the center of the image.
Figure 8 Rectangular structures south of Amenemhet III pyramid in VH-VV composite image (le\).
Reflec*on from high voltage tower A and three possible structures (B-D) delineated in SAR image (center)
and overlaid on Google Earth image (right).
Figure 9 False color images: a) Sen*nel-1 C-band SAR with VV and VH in green and red (le\), b)
ALOS/PALSAR L-band SAR with HH and HV in green and red (center), and c) fused principal components
image with first, second, and third principal components in red, green, and blue (right).

k-th color
combination average

Visible image SAR image Estimated SAR image

Figure 10 Nonlinear es*mator can be decomposed into a set of condi*onal averages over all k r-g-b
combina*ons of color values.
Figure 11 Nonlinear es*ma*on (from le\ to right): a) visible image, b) Sen*nel-1 VV polariza*on, c)
predicted Sen*nel-1 VV polariza*on from the visible image, and d) error image where red (green) are
posi*ve (nega*ve) errors where measured backscacer is greater (less) than the backscacer es*mated
from the visible image.
Figure 12 Topographic analysis. North-south transect (top) passes through the center of the Labyrinth.
The maximum eleva*on difference from north to south is –3 meters. West-east transect (bocom) passes
through the Abdul Wahbi canal. The maximum eleva*on difference from west to east is +12 meters.

Figure 13 View of the pyramid of Amenemhet III from the south along the Abdul Wahbi canal. (Image
courtesy Explore Faiyum19)

19
hcps://fayoumegypt.com/hawara-pyramid/