The Legendary Lost City Ciudad Blanca Found under Tropical Forest

in Honduras, using ERS-2 and JERS-1 SAR Imagery

Francis Yakam-Simen 1, Edmond Nezry 1, James Ewing 2
1 PRIVATEERS N.V. Private Experts in Remote Sensing
Great Bay Marina, P.O. Box 190, Philipsburg, Netherlands Antilles
Tel : (+33) 5-61545827, Fax : (+33) 5-61991724, Email : 106341.2602@compuserve.com
Internet Homepage : http://www.treemail.nl/privateers/
2 Ewing Data AB, Gotgatan 39, Stockholm, Sweden
Tel : (+46) 70-7410358, Fax : (+46) 8-6431106, Email : jim@ewingdata.com
Internet Homepage : http://www.ewingdata.com

Abstract
The legendary "Ciudad Blanca" of Honduras was first
referred to under the name Xucutaco by the Spanish
conquistador Hernn Corts already in 1526. Located in the
remote, impenetrable and incompletely mapped rainforest of
the Mosquito Coast, it was never conquered by the Spanish.
With the time, it was slowly abandoned and forgotten.
Two JERS-1 and one ERS-2 SLC Synthetic Aperture Radar
(SAR) images have been used to identify and to locate the
lost city, a task made difficult due to the thick vegetation
cover. To this end, advanced processing tools for the
detection of artificial targets under forest cover, and for SAR
data fusion have been used.
Among the techniques used, a new Bayesian Distribution
Entropy Maximum A Posteriori (DE-MAP) vector speckle
filter, particularly suited for the restoration of a strongly
textured scene, has been used to enhance the SAR images.
This new speckle filter incorporates a statistical description
of the effects of the SAR imaging system: in order to account
for the effects due to the spatial correlation of the speckle in
SAR images, an estimator originating from the local spatial
autocorrelation function (ACF) of the SAR signal are
incorporated to this filter, to refine the evaluation of the nonstationary first order local statistics, to improve the
restoration of the scene textural properties, and to preserve
the useful spatial resolution in the speckle filtered image.
On the other hand, radargrammetric techniques have
been used to: i) produce a Digital Elevation Model (DEM) of
the study area; ii) fuse ERS and JERS information in order
to allow visual identification of the remnants of Ciudad
Blanca by visual photo-interpretation.
Using the processed images, geocoded UTM spatio-maps
of the region have also been produced to locate accurately
our findings, and guide a ground expedition in the future.

1.

Introduction

In the eve of the 16th century, a great trading

Mayan/Nahua city was flourishing in what is known today
as the Mosquito Coast, in North-Eastern Honduras. This city
was known by the native Central American peoples, from as
far as Panama and Mexico. This city was first referred to
under the double name of Xucutaco (Nahuat) and
Hueitapalan (Mayan) by Hernn Corts already in 1526 [1].
Corts himself affirms that he was informed of the existence
of this city shortly after he reached the coast of the Gulf of
Mexico in 1519. After Corts was told and reports in [1],
Xucutaco-Hueitapalan was an impressive city, comparable
in population and in wealth with Mexico/Tenochtitlan itself.

After Hernn Corts had to give up heading to it, the city

protected by the impenetrable rainforest of the "La
Mosquitia" was ignored by the Spanish conquistadores. With
time, Xucutaco-Hueitapalan was left to the jungle by its
inhabitants, apparently already in the middle of the 16th
century. Nevertheless, indigenous people still remembered it
as the legendary White City ("Ciudad Blanca").
In 1960 Honduras created the Ciudad Blanca
Archaelogical Reserve (ca. 5250 km) for the Pltano River
region and in 1969 made it an archaeological national park.
This recognition endorsed the importance of archaeological
finds and legends of an ancient major Maya city in the area,
although actual protection was minimal [2]. The cultural
significance led to other scientific evaluations, confirming
that the Mosquitia region of Honduras and Nicaragua had
the most virtually undisturbed forest of northern Central
America. As a result, in 1980 the park was recognized as the
Ro Pltano Biosphere Reserve (BR), the first in Central
America, under UNESCO's Man and the Biosphere
Programme (MAB), which provided stronger legal protection
and a management planning process for the Pltano River
watershed and adjacent slopes. In 1982 the BR was
accepted on the World Heritage List [3].
The region has long been a site of human occupation. It
has many archaeological sites (e.g. Maraones, Lancetillal,
Platanillales, Saguasn, Limeta), including mysterious
petroglyphs carved into large boulders along the river edge.
The "Instituto Hondureo de Antropologa e Historia" (IHAH)
is conducting studies in the BR. The village Las Crucitas del
Ro Aner in the reserve's south-east (on the Aner River a few
kms from the Wamp River) is established over one of the
largest and most impressive archaeological sites [4]. It is
believed that the fabled Ciudad Blanca awaits discovery
within the reserve. Additional archaeological research in the
region may confirm the surmise that its peoples were an
important link between major pre-Columbian cultures in
North America and South America.
The project to accurately identify and locate the the ruins
of Ciudad Blanca has been initiated in November 1997, with
the aim to contribute to return to Honduras and to mankind
a bright page of their history, and to put the site under
protection of the due authorities. To do so, the project was
carried out under the auspices of SEPHA, the "Sociedad
para la Exploracin y la Proteccin de la Historia de las
Americas", Honduras, between December 1997 and
February 1998. The remote sensing part of the project has
been kindly supported by the European Space Agency
(ESA/ESRIN) in Italy, and by the National Space
Development Agency of Japan (NASDA) who generously
provided the satellite remote sensing data needed.

Identification and location of the lost city has been carried

out using two JERS-1 3-looks (Level 2.1) amplitude SAR
images and one ERS-2 SLC (Single Look Complex) SAR
image. As expected, such a task was made challenging by
the presence of very thick vegetation, since the radar wave
does not penetrate the whole vegetation cover of a tropical
forest, even at L-band.
Such a target was not expected to be directly visible by
spaceborne SAR. However, the disturbances its presence
causes to the natural arrangement of vegetation was
expected to be somehow detectable in SAR images. To reach
our goal new SAR image enhancement techniques had to be
developed.
2.

SAR Images Enhancement and the DE-MAP Speckle

Filter
In the case of a single-channel detected SAR image, let
define the quantities of interest: I is the speckled intensity
available in the actual SAR data; R is the radar reflectivity
which is the quantity we will to restore. The Maximum A
Posteriori (MAP) filtering method bases on the famous Bayes
theorem (1828):
(1)

P(R/I) = P(I/R) P(R)/P(I)

The MAP solution is obtained when the posterior probability

is maximum, i.e.:
Ln(P(I/R))/R + Ln(P(R))/R = 0

for R = R MAP (2)

The density probability function P(I/R) of the speckle in the

L-looks SAR image can be modelled as a Gamma
distribution:
P(I/R) = (L/R)L /G (L) exp( -L I/R) I (L-1)

(3)

In presence of very strong texture, which is the case in our

images with ancient ruins in a very dense tropical forest and
in the presence of relief, it may be hazardous to make an
assumption about the probability density function of the
radar reflectivity. In this context, the A Priori knowledge with
regard to the observed scene can hardly be an analytical
first order statistical model. However, we introduce a
Maximum Entropy constraint on texture [5-6]:
E(R) = S = - [Rk log(Rk ) ]

(4)

LI - LR - R 2 [log(Rk ) - 1/Ln(10)] = 0

(6)

Due to the nature of the targets we are interested in, and to

the nature of their environment, this new speckle filter is
particularly suited as an image enhancement tool in the
present application.
To comply with the basic assumptions of stationarity and
ergodicity, the second order statistical spatial properties
(spatial correlation) of all the individual SAR images must
also be considered to include both scene texture and
resolution/sampling related properties, for an optimal
restoration of the radar reflectivity. In addition, a
preliminary detection and identification of the structural
elements of the scene is required.
To cope with these requirements, locally estimated twopoints statistics have been introduced in [7-9] to perform
structure detection as well as to refine the computation of
the local Non-stationary Mean Non-stationary Variance first
order statistics. This detection algorithm is inspired from the
concept of "texture fields" [10]. This concept, based on the
analysis of second order statistical properties, generalises
image processing concepts such as filtering and detection,
which are usually considered as distinct. It enables to
perform simultaneously structure detection and image
filtering.
Regarding speckle filtering, the local mean radar
reflectivity estimation takes then into account the textural
properties of the imaged scene (spatial variability of R
between resolution cells, directionality) and the spatial
correlation properties of the speckle i.e. the characteristics
of the point spread function of the imaging system.
Introducing the estimates for the non-stationary local
statistics into the scalar equation of a single-point filter
improves the restoration of natural R fluctuations (texture)
in the filtered SAR image, as it has already been shown in
Nezry et al. (1995) [7]. Moreover, it is mandatory for the
estimation of the Rks in the DE-MAP filter.
In addition, using this technique, the detection of even
thin or low-contrast structures becomes
implicit.
Consequently, a specific structure detection associated to
the filter is no longer needed, and the overall processing is
greatly simplified. The major advantage is that there is no
geometrization of the scene structural elements, as it may
happen using the templates-based Ratio-Of-Amplitudes
(RoA) detectors [11].

Because the Rk are non-negative and 1/Z. exp(S(R)) is

normalized, it can be treated as a probability distribution
whose entropy is S(R) [5]:

(5)
P(R) = 1/Z exp(S(R)) = 1/Z exp - [Rk log(Rk ) ]
k

To estimate P(R), the radar reflectivities Rk are estimated

using the process of deconvolution from the SAR impulse
response function which is described in [7].
Under these assumptions, the Distribution Entropy Maximum
A Posteriori (DE-MAP) filter for single-channel detected SAR
images is the solution of the following equation:

3.

Processing of the ERS-2 and JERS-1 SAR Images

The available SAR images are: an ERS-2 Single Look

Complex (SLC) C-VV band SAR image (acquisition: 25
October 1996), and two 3-looks JERS-1 (Level 2.1) L-HH
band SAR images (acquisitions: 06 February 1995 and 18
July 1996).
Pre-processing of the remote sensing images has been
particularly stressed. Indeed, the quality of this preprocessing contributes substantially to the accuracy of the
final product.

Complete radiometric calibration of the SAR images is

carried out, according to the ERS and JERS respective
specifications.

The ERS-2 SLC image is multi-looked in azimuth to

obtain a 4-looks amplitude image. This amplitude image
is then resampled from slant-range to ground range,
with a final pixel size of 20x20 m.
The JERS-1 images are resampled (linear resampling)
to 20x20 m.
Then speckle noise filtering is performed. To this end,
the new DE-MAP adaptive speckle filter was used.
Detectors based on the local spatial autocorrelation
functions of both the speckle and the scene in SAR
images are incorporated to these filters. They improve
the restoration of the scene textural properties and of the
scene structural elements, to produce speckle filtered
images without loss in spatial resolution.
First of all, a SAR stereogram is produced, using the
ERS-2 (incidence angle ranging from 23.67 and 26.51)
and JERS-1 (incidence angle ranging from 35.12 and
31.27) SAR images. From this stereogram, a Digital
Elevation Model (DEM) is produced by radargrammetry,
with a pixel size of 20x20 m (cf. Figure 2). In the present
study, the horizontal DEM accuracy is better than 15
meters; the vertical DEM accuracy is better than 20
meters.
In remote sensing images acquired by side-looking
sensors, relief is systematically displaced with regard to
its true geographical location, due to the viewing
geometry. This effect is particularly important in SAR
images. Thus, the local geometrical corrections
(geocoding, or in other words ortho-rectification) are
performed, using the DEM.
The radiometric corrections for slope angle and aspect
angle effects are carried out, using the DEM. These
corrections account for the variation of effective
scattering area with local slope and orientation, as well
as for the variations in illumination conditions. Effects of
the variations in illumination conditions on the image
radiometry are evaluated using a physical backscattering
model

4.

Figure 1. UTM (WGS84) map of the La Mosquitia area around the site of
Ciudad Blanca. The area shown is 59 x 81 km.

Cartography of the Site (La Mosquitia, Honduras)

Once all SAR images have been filtered, geocoded and

radiometrically corrected, they are georeferenced into
Universal Transverse Mercator (UTM) projection, using Earth
ellipsoid WGS 1984.
At this point, a spatio-map is produced. The grey levels in
this map are obtained by Bayesian fusion of the filtered,
geocoded, radiometrically corrected and geo-referenced SAR
images [9].

5.

Ciudad Blanca Unveiled

First of all, we tried to identify already known and

documented sites in our SAR dataset. As an example, the
site known as Las Crucitas I [4] is shown in Figures 3 and
4 below. This site is one of the many archaeological sites
that compose Ciudad Blanca. The stereoscopic ERS2/JERS-1 unfiltered SAR images are shown in Figures 3.
The stereoscopic ERS-2/JERS-1 enhanced images are
shown in Figure 4. The features that are visible on this
stereoscopic image correspond satisfactorily with the map of
the site published by Lara-Pinto and Hassemann [4].

Figure 2. DEM obtained by ERS-2/JERS-1 radar-grammetry. The area

shown is 59 x 81 km.

Figure 3. The already documented site of Las Crucitas I (Ciudad

Blanca). Unfiltered SAR data. Area size: 2x2 km. (Left: ERS; right: JERS1)

Figure 6. The great undocumented site found (Ciudad Blanca. DE-MAP

Filtered. Area size is 2x2 km. (Left: ERS; right: JERS-1).

It is noticeable that the results of a systematic

bibliographic research conducted in Honduras, in the USA
and in Europe from January 1998 to May 1998 show that
this site has not been documented until now. It has
probably not even been explored recently (at least by writers
or scientists ...). The last report of a place structurally
similar to this one in La Mosquitia dates back to 1544, when
a Spanish missionary was guided there by local people.

Figure 4. The already documented site of Las Crucitas I (Ciudad

Blanca). DE-MAP Filtered. Area size: 2x2 km. (Left: ERS; right: JERS-1)

Similarly, a number of other already documented sites, as

well as numerous minor artefacts that are probable ruins
scattered in a ca. 20x20 km wide area near have been
identified by a careful examination of the processed images.
Nevertheless, an important finding was made during this
systematic examination. Covering a 3.0x3.5 km wide area in
one of the denser part of the forest (and close enough to Las
Crucitas to be connected to the sites that are already
documented there), the ruins of a vast complex of important
structures are visible in the images. The Figures 5 and 6
illustrate the most interesting part of this area, including
what is very probably a vast ceremonial center (Upper left
quadrant of the images). The stereoscopic ERS-2/JERS-1
unfiltered SAR images are shown in Figures 5. The
stereoscopic ERS-2/JERS-1 enhanced images are shown in
Figure 6.

Figure 7. Map of the probable ceremonial center found in this study.

Figure 7 is an enlargement of the important structures

located in the upper-left quadrant of Figures 5 and 6. This
enlargement has been made by photo-interpretation of these
images.
From the elevation map (DEM), and the shadows observed
in the images, one may infer the presence of a pyramid (or a
high square-shaped structure) in the north-western part of
the area represented in Figure 7.

6.

Figure 5. The great undocumented site found (Ciudad Blanca).

Unfiltered SAR data. Area size: 2x2 km. (Left: ERS; right: JERS-1).

Conclusions

The legendary lost city Ciudad Blanca has been identified

in ERS-2 and JERS-1 SAR data. Its ruins are scattered over
a very wide area (ca. 20x20 km) in La Mosquitia, Honduras.
Some major centers had already been found and
documented by archaeologists. Nevertheless, an important
still undocumented complex of ruins, which is apparently
the most important in the area, has been discovered using
ERS-2 and JERS-1 SAR images, enhanced by means of the
new DE-MAP speckle filter developed by Privateers NV.
In addition, 6240 km2 have been mapped around Ciudad
Blanca, in La Mosquitia, Honduras. In this area, still

incompletely mapped, this map will be useful to guide a

forthcoming expedition to document the findings.
It is noteworthy that, although C-band SAR as ERS is
generally not expected to detect structures under dense
tropical forest, it has proven here as efficient as L-band SAR
in our search for Ciudad Blanca. Given the density of the
rainforest in La Mosquitia, it is anyway unlikely that even
the JERS-1 L-band waves could completely penetrate
through the forest cover. In fact, we believe that SAR
detection of the ruins is indirect: what we detect correspond
merely are the perturbations caused by the ruins to the
structure and to the density of the forest cover.
Note that the accurate coordinates of the new major site
found in this study are, to the moment, reserved to the
Instituto Hondureo de Antropologa e Historia (IHAH), in
order to preserve the site.

7.

Aknowledgement

The ERS-2 image has been provided by the European

Space Agency to PRIVATEERS N.V. in the framework of the
ERS Pilot Project PE-FRNE2. The JERS-1 SAR images have
been provided by NASDA, the National Space Development
Agency of Japan, to PRIVATEERS N.V. in the framework of
the JERS-1 Research Project #675. In particular, a great
deal of gratitude is due to NASDA, Japan and to Dr. Henri
Laur (ESA/ESRIN, Italy), as well as to RESTEC, Remote
Sensing Technology Center of Japan, for their generous
support and as a recognition of their width of view.
Original ERS-2 images: ESA/EURIMAGE 1996
Original JERS-1 images: NASDA 1996; MITI/NASDA
retains ownership of JERS-1 data; NASDA supports
PRIVATEERS N.V. in acquiring the satellite data.

8. References
[1] H. Corts, 1526: "Cartas de relacin de la conquista de
Mxico", 5th letter to Emperor Charles V, Tenuxtitn, 3rd
September 1526.
[2] WWF-US, 1988: Country plan-Honduras. In: "1989/90
program: Latin America and the Caribbean: new and
ongoing projects", World Wildlife Fund-US (WWF-US),
Washington D.C., pp.291-312.
[3] B. Houseal, C. MacFarland, G. Archibold and A. Chiari,
1985: "Indigenous cultures and protected areas in Central
America", Cultural Survival Quarterly, 9 (1): 10-20.
[4] G. Lara-Pinto and G. Hassemann, 1991: "Leyendas y
arqueologa: Cuntas ciudades blancas hay en la Mosquitia
" In Murphy, V. (ed.), "La Reserva de la Bisfera del Ro
Pltano", Ventanas Tropicales, Tegucigalpa (Honduras). pp.
16-19.
[5] A.K. Jain, 1989: "Fundamentals of digital image
processing", Chapter 8, Prentice Hall Ed., Englewood Cliffs
(NJ), 1989.
[6] M. Datcu and M. Walessa, 1997: "Maximum entropy
methods for despeckleling and resampling of synthetic
aperture radar images of rough terrain", Proceedings of
SPIE, 3917, pp.76-83.
[7] E. Nezry, M. Leysen and G. De Grandi, 1995: "Speckle
and scene spatial statistical estimators for SAR image
filtering and texture analysis: some applications to
agriculture, forestry, and point targets detection",
Proceedings of SPIE, 2584, pp.110-120.
[8] E. Nezry, F. Zagolski, A. Lops and F. Yakam-Simen,
1996: "Bayesian filtering of multi-channel SAR images for
detection of thin structures and data fusion", Proceedings
of SPIE, 2958, pp.130-139.
[9] E. Nezry, F. Yakam-Simen, F. Zagolski and I. Supit,
1997: "Control systems principles applied to speckle
filtering and geophysical information extraction in multichannel SAR images", Proceedings of SPIE, 3917, pp.48-57.
[10] O.D. Faugeras and W.K. Pratt, 1980: "Decorrelation
methods of texture feature extraction", IEEE Transactions
on Pattern Analysis and Machine Intelligence, PAMI-2, n4,
pp.323-332.
[11] A. Lops, E. Nezry, R. Touzi and H. Laur, 1993:
"Structure detection and statistical adaptive speckle
filtering in SAR images", International Journal of Remote
Sensing, 14, n9, pp.1735-1758.