Month: June 2013

Archaeological Prospection: Japanese Satellite

Kids Talk Radio Japan: Many nations have satellites traveling around the earth. Japan is a one of the leaders in this technology. Students in the STEAM Plus programs are studying how satellites are being used today. Our goal is to gain a deeper understanding for our school of
the future projects in Long Beach, California (USA).


The work presented here is a further study of the UNESCO Cultural Heritage sites of Samarra (Iraq) and Djebel Barkal archaeological area (Sudan) by means of polarimetric products of the Japanese satellite ALOS PALSAR. Despite the Iraqi war ending in 2011, the city of Samarra is still included on the UNESCO List of Sites in Danger (since 2007). The study of the city presented here began three years ago with the analysis of optical data. The work showed an urban and agricultural expansion affecting the integrity of the city. An attempt to study Samarra by using the polarimetric SAR technique was made in previous years with the employment of the ALOS PALSAR satellite. That study was limited to the analysis of polarimetric descriptors such as entropy and alpha angle, parameters that gave results regarding prevalence of double-bounce and volume scattering mechanisms in the area investigated. The second archaeological site, Djebel Barkal, is one of five archaeological sites located in a semi-desert area along the River Nile, in the Napatan Region considered to be part of Nubia. The site was included in the UNESCO World Heritage List in 2003. Polarimetric descriptors considered in previous studies of the zone were entropy and alpha angle. Also in this case, a deeper analysis was carried out with the addition of a second ALOS PALSAR polarimetric SAR image, acquired 3 years later. In the present work, more polarimetric parameters, such as Freeman and Yamaguchi decompositions, are taken into account in order to observe scattering mechanisms both upon the structures already known and the area around them. Thanks to these decompositions and the archaeological maps that were available, it was possible to validate backscattered responses in ALOS PALSAR images as archaeological structures. Copyright © 2013 John Wiley & Sons, Ltd.


The recognized usefulness of remote observation for the analysis of archaeological structures started at the end of the nineteenth century when in 1879 the ancient city of Persepolis was studied using aerial photographic documentation. From the 1890s until the beginning of World War II, a new way of documentation for sites in desert areas was inaugurated with airborne acquisitions over some Near Eastern ancient sites, such as the city of Troy, Roman cities in Palestine, in Macedonia and as far as Syria and Mesopotamia (Piccarreta and Ceraudo, 2000). The arrival of optical satellites with an appropriate spatial resolution allowed the detection of underground ancient structures, thus reducing aerial campaign costs specifically planned for archaeological purposes (Lasaponara and Masini, 2012).

As their usefulness has been well established for several years, our attention focused on the SAR polarimetric technique as a new way of investigation for cultural heritage applications. The reasons linked to our approach on SAR sensors are due, first, to the 24-h observation independent of Sun illumination and cloud coverage, and to the importance of providing additional information concerning electromagnetic properties of the targets, qualities not obtainable from optical images.
Test sites

The sites taken into account as test areas are two important ancient cultural heritage sites under the aegis of the United Nations Educational Scientific and Cultural Organization (UNESCO, 1992).
Samarra threats, climate and geology

Samarra archaeological area (N43°45’–43°51′, E34°25’–34°05′) was inscribed in 2007 in the list of Cultural Heritage of sites in Danger. One major threat for the safeguard of the property arises from the inability of the responsible authorities to exercise control over the management and conservation of the site, as protective procedures have been in abeyance since 2003. Prior to recent hostilities, the State Party protected the site from intrusions, whether farming or urban, under Archaeological Law (UNESCO, 2007).

The reason why the ancient capital city of Iraq was selected as a test site is linked to two important points: its low level of precipitation during the year and to its advantageous geological situation.

Samarra has the typical semi-desert climate of central Iraq. It is an extremely dry continental version of the Mediterranean climate, characterized by sparse rainfall falling during winter months, and high temperatures in the summer. Rainfall at Samarra results from depressions tracking across the Mediterranean, and crossing the Syrian Desert before depositing rain on Iraq (Northedge, 2007).

The city is located north of the limit of the alluvial plain on the edge of the Tigris valley, where the river is incised between 15 and 20 m into the rolling steppe east and west of the river, and the floodplain is 2–4 km wide. The River Tigris dominates the geology of the Samarra region. The steppe, on both the sides of the river, is composed of three fluvial terraces formed by deposits of the Rivers Tigris and Euphrates during the Pleistocene era, with large numbers of gravel-sized clasts present in the matrix. In the 1950s, the terraces of the Samarra region were given the names Mutawakkil, Mu‘tasim and Mahdy terraces. At Samarra, the Mutawakkil terrace is dominant. It is composed of rounded clasts of chert, limestone and metamorphic rocks in a clayey and sandy matrix. This material is relatively easy to erode by water or wind action and, in addition, it is relatively easy to excavate for the construction of buildings (Buringh, 1960).
Samarra history

Samarra was the second capital of the Abbasid Caliphate. Capital of Iraq only for 56 years (ad 836–892), when the city was suddenly and mysteriously abandoned. This allowed very good preservation of its topography and provides an important example for the study and comprehension of the ancient urban system for archaeologists. The portion of the whole area of Samarra properties (15 058 ha) presented in this work is the octagonal city of al-Qadisiyya, an unfinished city built in mud-brick and still unexcavated. It was built, according to the texts, by Caliph Harun al-Rashid on the model of the Round City of Baghdad and abandoned in ad 796, before Samarra city was built (Northedge, 2007).
Djebel Barkal threats, climate and state of conservation

The archaeological area of Djebel Barkal is located between the cultivable west bank of the Nile river in Sudan and the semi-desert sandstone rock outcrops of the Sudan desert (N18°32′, E31°49′). The area is very close to the cultivated fields of the city of Karima, and their number increases year on year. Being protected by UNESCO, the integrity of the archaeological structures has been maintained; large modern infrastructure remains outside of a buffer zone, only partially established, but close to its limits. At present, the archaeological structures are only very slightly affected by modern urban spread. However, careful monitoring of the developments around the property needs to be carried out, especially urban extension on the desert side of Karima. The property is guarded by a military force from the Police of Tourism and Antiquities.

The absence of heavy or continuous rain phenomena is the major reason why the nine temples were discovered around the 100 m high sacred mountain of Djebel Barkal and its pyramid complexes. The latter are made from stone and still present their original shape and internal division, as well as burial chambers and wall paintings. This is not the case for many of the palaces, which are made of earthen sun-dried bricks (UNESCO, 2003).
Djebel Barkal history

Djebel Barkal is one of five Napatan (900 to 270 bc) and Meroitic (270 bc to ad 350) archaeological sites stretching over more than 60 km in the Nile valley, in a semi-desert area considered to be part of Nubia. It was the capital of the Kushite kingdom, probably by the end of ninth century bc, keeping both its religious and administrative roles until the fourth century (Kendall, 2002). This area was considered to be sacred, with its tombs, temples and pyramids, since New Kingdom times (ca. 1500 bc). Several temples at the foot of the sacred hill and facing the Nile, have been revealed by excavations and survey, together with administrative palaces and pyramids.

The pyramids and tombs, being also part of the special desert border landscape, on the banks of the Nile, are unique in their typology and building technique. They represent part of the necropolis of the Napatan–Meroitic cemetery. The Napatan–Meroitic pyramids are built differently and are conceptually dissimilar compared with the Egyptian examples. Whereas the Egyptian models were built to enclose and hide the burial chamber, the Napatan ones were commemorative monuments to the dead, buried in a hypogeum underneath. In front of the pyramids, small temples were built for offerings. The 30 tombs explored are accessible and most of them are decorated, with either paintings or engravings. The site has vast archaeological areas that have neither been excavated nor studied. Generally, archaeological excavations at Gebel Barkal have not reached the earliest strata. Thanks to the special cultural aspect of the site and to its geomorphological characteristics, the analysis of SAR polarimetric data was focused on the northwest and the central groups of Royal pyramids and their surroundings.

For both the sites, the datasets are composed of ALOS PALSAR polarimetric SAR images, meteorological archived information, optical images and archaeological maps; these two last datasets were used as part of the interpretation rationale.
Samarra documentation

The Samarra optical dataset comprises two ALOS PRISM images acquired on 11 April 2009 and 22 November 2009. Due to the wide extent of the area of interest (north–south length of about 41.5 km, with a width varying from 4 km to 8 km), mosaic of these two images was made in order to have the whole city in one image. This mosaic was then used for the georeferencing process applied to all the SAR products. The archaeological maps used as tests for product interpretation are available on the UNESCO website (UNESCO, 2006). These maps comprise satellite imagery dated July 2004, January 2005 and March 2006, with a resolution of about 60 cm (Digital Globe), and Landsat Enhanced Thematic Mapper imagery dated approximately 2000, with a resolution of about 15 m. Overlay maps were created by Alastair Northedge, Director of the Samarra Archaeological Survey. These maps were used as a validation system both for the known structures and as support for the interpretation of the surrounding areas of al-Qadisiyya, where structures are not visible in optical images. Cartographic documentation was georeferenced in the Universal Transversal Mercator projection (WGS 84, zone 38N) using as a base the optical ALOS PRISM mosaic. The only SAR image that we used was acquired on 15 November 2008 and is available via the European Space Agency (ESA) on line catalogue EOLI-SA (ESA, 2009). It was acquired by ALOS satellite PALSAR sensor in polarimetric mode with an incidence angle of 23.10°. Given that the Satellite activity ended in April 2011, we were unable to obtain additional images over the area. As a result multitemporal analysis and comparison between different images have not been performed.
Djebel Barkal documentation

The optical data set we used at Djebel Barkal is one KOMPSAT-2 image acquired on 16 May 2008, used as a validation system and as a basis for the georeferencing process (Universal Transversal Mercator projection, WGS 84, zone 35N). Concerning the cartographic documentation, we used only one archived archaeological map derived from UNESCO reports over the area (UNESCO, 2011). The two full-polarimetric archived ALOS PALSAR images analysed in this work are available on the ESA online catalogue (ESA, 2009). They were acquired with a time interval of 3 years, the first one on 14 August 2006 and the second one on 5 November 2009, with an incidence angle configuration mode of 26.20° and 23.10° respectively. Thanks to this time interval, a multitemporal comparison between the data, even if not the same season of the year, was carried out. The multitemporal analysis demonstrates how in three years many of the modern buildings surrounding the UNESCO property are built in the area. Unfortunately, we do not have further optical data, neither satellite nor aerial, or further cartographic documentation that could extend the analysis.

Polarimetric SAR processing is carried out by means of PolSARpro software. This tool is dedicated to the accessibility and exploitation of multipolarized SAR datasets. PolSARpro is a software developed under contract to ESA by a consortium comprising IETR at the University of Rennes 1, the Microwaves and Radar Institute (HR) of Deutsches Zentrum für Luft- und Raumfahrt e.V (DLR) and AEL Consultants, together with Dr Mark Williams. By means of this software it is possible to analyse different spaceborne and airborne sensors (ESA, 2011).

The ALOS PALSAR data requested have a 1.1 processing level, range and azimuth compressed. It means that with this level the image is a slant range complex data.

Pre-processing consists in identifying the areas of interest within the image, performing a multilook operation with five pixels in each row and one in each column, in order to obtain a squared pixel size. This is a good compromise between a loss of spatial resolution and an unbiased estimation of the polarimetric parameters when performing polarimetric decomposition.

A Pauli RGB image is derived automatically from the multilook data and it is usually used as a beginning step to interpret the physical information from a qualitative point of view (Patruno et al., 2009).

The first step for these polarimetric data concerns the extraction of the T3 matrix. Polarimetric descriptors analysed in a previous research are entropy (H) and alpha angle (α). Entropy provides information about the randomness of the scattering mechanism, while α is important in order to evaluate the predominant scattering mechanism between single-bounce, double-bounce or volume scattering (Lee and Pottier, 2009). In Figure 1 an example of entropy and alpha angle from the al-Qadisiyya area together with H/α plane is shown. As visible, high values of H and medium values of α are recorded in almost the whole of the zone corresponding to surface roughness (Zone 8) and vegetation (Zone 5) in the H/α plane. Except for part of the octagonal city walls and part of interior system of the canals (black ellipses in the image), both recorded with low values of H and low values of α, no other archaeological features were detected by the ALOS PALSAR sensor (Dore et al., 2010).

Figure 1. Entropy (left) and alpha angle (right) polarimetric products. Blue colour indicates low values, green colour medium values and red colour high values, as indicated in the colour key under each image. Below the images is the H/α plane.

The low archaeological information gained with H and α polarimetric descriptors turned our attention to other parameters in order to test the validity of the scattering mechanisms both upon the already known structures and in the areas around them.

The products we analysed are the Freeman and the Yamaguchi decompositions. The Freeman decomposition proposed by Freeman and Durden (Freeman and Durden, 1998; Freeman, 2007) is an important model-based incoherent decomposition. It models the polarimetric coherency matrix as the contribution of three scattering mechanisms: surface, double-bounce, and volume scatterings (Figure 2). It is a technique for fitting a physically based, three-component scattering mechanism model to the polarimetric SAR observations, without ground-truth measurements. It can be used to describe polarimetric backscatter from naturally occurring scatterers. This three-component scattering power model can be applied successfully to decompose SAR observations under the reflection symmetry condition (Nghiem et al., 1992). The mechanisms taken into account are canopy scatter from a cloud of randomly oriented dipoles, even or double-bounce scatter from a pair of orthogonal surfaces with different dielectric constants and Bragg scatter from a moderately rough surface. The first component of Freeman decomposition consists of a first-order Bragg surface scattering, modelling slightly rough surface scattering. The double-bounce scattering component is modelled by scattering from a dihedral corner reflector (e.g. ground–tree trunk backscatter) where the reflector surface can be made of different dielectric materials. The volume scattering is modelled as the contribution from a cloud of randomly oriented cylinder-like scatterers (Lee and Pottier, 2009).

Figure 2. The three images represent the three scattering mechanisms of Freeman decomposition associated with RGB colours. Red (left image), Freeman_DBL; green (central image), Freeman_VOL; blue (right image), Freeman_ODD.

As mentioned above, Freeman decomposition produces its best results in reflection on azimuthal symmetry conditions. Scattering symmetry assumptions about the distribution of the scatterers lead to a simplification of the scattering problem and allow quantitative conclusions about their scattering behaviour. If the scattering matrix S for a target is known, then the scattering matrix of its mirrored or rotated image in certain symmetrical configurations can be derived immediately. Considering a distributed target that has reflection symmetry in the plane normal to the line-of-sight (Figure 3), whenever there is a contribution from a point P there will always be a corresponding contribution from its image at point Q (Nghiem et al., 1992).

Figure 3. Reflection symmetry about the line-of-sight (after Lee and Pottier, 2009).

Considering urban areas, where symmetry conditions are not valid, a Yamaguchi three-component decomposition is performed (Lee and Pottier, 2009). This decomposition starts from the assumption that in nature, reflection with symmetric conditions is not so common.

Both decomposition procedures produce a RGB image where the red channel corresponds to the double-bounce scattering, the green channel corresponds to volume scattering and the blue channel corresponds to single-bounce scattering. This allows an interpretation of the physics behind the colours represented.
Samarra results

Figure 4 shows the Freeman decomposition. The image covers the southern part of the Samarra archaeological area (al-Qadisiyya city) selected for SAR analysis and validation. What can be detected in the image, from an archaeological point of view, are features also well known in optical images and previously recorded in the literature. The wall perimeter is well detected and visible in the Freeman image (white arrows). It responds to a single-bounce scattering mechanism. This type of response, instead of a double bounce, is due to the angle created between wall inclination and radar incidence angle (23.10°). Other interior channels are also well visible and, for them as well, the response is a single bounce. What is notable is that there are no responses from the qanāt, the underground channel entering the city through the northwest wall with clear surface alterations. At a first level of interpretation we may ascribed this absence of response to the low ALOS PALSAR spatial resolution (about 20 m), but we realized that it cannot be possible as the width of the surface alterations is about 64 m wide. Most of all, it is possible to detect the qanāt in ENVISAT-ASAR imagery (a series of images spanning from 2004 to 2008) with a spatial resolution of about 30 m. These results therefore are not due to their spatial resolution, but to the deformation of the surface and that the C-band is more sensitive to it.

Figure 4. Freeman decomposition. White arrows indicate the single-bounce scattering mechanism of the perimeter of the octagonal city. White ellipse indicates the area where the qanāt is located (R, double bounce; G, volume scattering; B, single bounce).

In the Yamaguchi image (Figure 5) the same observations can be made as for the Freeman decomposition.

Figure 5. Yamaguchi 3 decomposition. White arrows indicate the single-bounce scattering mechanism of the perimeter of the octagonal city. White ellipse indicates the area where the qanāt is located (red, double bounce; G, volume scattering; B, single bounce).
Djebel Barkal results

Concerning Djebel Barkal, the same polarimetric decompositions were performed. The area investigated is the portion northwest of the archaeological complex, near the first group of Royal pyramids. Here, a strong backscatter is noticed in all the products obtained: Pauli RGB (Patruno et al., 2013), Freeman and Yamaguchi decompositions.

In the Freeman decomposition performed over the ALOS PALSAR image acquired on 14 August 2006 (Figure 6), it is possible to notice a very bright response to the radar waves (yellow ellipse), that is confirmed also in the Freeman decomposition of the 2009 image (Figure 7).

Figure 6. Freeman decomposition. 2006 image. The yellow ellipse outlines the strong backscatter close to the northwest group of Royal Pyramids (R, double bounce; G, volume scattering; B, single bounce).

Figure 7. Freeman decomposition. 2009 image. In the yellow ellipse, the strong backscatter is still persistent after 3 years (R, double bounce; G, volume scattering; B, single bounce).

Similar responses have been analysed all over the archaeological area in order to understand if the morphology of the site, made of sandstone rocks, could provide the same backscattering mechanism. Observing the two ALOS PALSAR images we notice several similar backscattering features caused by geology, but never in the same place and never corresponding to a determinate pixel in the both images. This led us to conclude that only for that particular bright response is there a correspondence in both images. Moreover, the 3 years difference in time acquisition is an important factor when considering that we can see this backscattering in the same place after 3 years.

As this point the target is very visible in an area where a high content of archaeological structures is present, but it is not visible in optical images (Figure 8). This could be due to penetration of radar waves. A wave penetration could be possible due to the PALSAR sensor L-band, and also to the incidence angle acquisition mode of PALSAR images (26° and 23°). The detection of a strong point target, probably underground, is confirmed also by Yamaguchi decomposition (Figures 9 and 10).

Figure 8. KOMPSAT-2 image. In the ellipse is the area where the anomaly was detected in Figures 6 and 7. In rectangles, are the three archaeological areas where remains are already known.

Figure 9. Yamaguchi 3 decomposition. 2006 image. The yellow ellipse outlines the persistence of the strong scattering noticed in the area (R, double bounce; G, volume scattering; B, single bounce).

Figure 10. Yamaguchi3 decomposition. 2009 image. The persistence of the strong scattering noticed in the area in 2006 is still clearly detectable in the yellow ellipse (R, double bounce; G, volume scattering; B, single bounce).

The application of a polarimetric analysis for archaeological purposes needs more confirmation via test sites, especially in desert, semi-desert or arid areas. This is not just to have a higher probability of radar wave penetration into the ground, but also to derive a possible validation of the method for the archaeological research. In this sense, the methodology, if recognized as valid, could have a great scientific potential, mostly for the areas of the world with a special cultural interest but which are exposed to natural, urban or agricultural threats. A further element to be considered is the opportunity to monitor cultural heritage areas where, due to instable political conditions, a validation in situ is not possible.

Moreover, the cited technical characteristics that SAR sensors have (independence from an external source of illumination, penetration of cloud coverage, wave penetration of the ground – according to the wavelength – and derivation of dielectric properties of the target) can extend the limits often encountered with optical satellite acquisitions. This is true even if optical satellites still have a better spatial resolution for archaeological observations. Indeed, the combined use of both sensors constitutes the best remote sensing technique for this field of research.

The two sites of Samarra and Djebel Barkal were selected mainly for their environmental conditions and for structures large enough to be identified by the PALSAR sensor. In this way it has been possible to detect structures already known and recorded in past years. In the case of Samarra, we concluded that, making an exception for features less than 20 m in extent, results are not only due to spatial resolution of the sensor, but also to the deformation of the surface. At Djebel Barkal, the identification of a strong backscatter, confirmed in both SAR images, is the most important contribution from the SAR image. The persistence during the years of the strong response allow us to suggest the permanent presence of a target, in the area of the northwest group of Royal pyramids. However, in order to confirm or discard the hypothesis of a microwave penetration, a DSM (digital surface model) will be created to test if the terrain morphology has a stronger contribution in that area instead of a radar penetration.

Surveys in situ are increasingly necessary for method validation. At the moment, unstable political situations do not allow ground-truth monitoring of the zones for this study. If the internal situations in the areas investigated do not change, then further satellite data would be useful to remotely observe the state of conservation of the sites.

The technique presented here started as a topic of a Masters degree in Archaeology at the University of Rome ‘Sapienza’ jointly with the use of optical data as a validation system. This topic was then developed as a Doctoral project by the Faculty of Engineering (DICEA Department) at the ‘Sapienza’ University of Rome. This brought a collaboration between the Italian University and the French University of Rennes 1 (IETR Department) carrying on a collaboration with the Italian establishment (ESRIN) of the European Space Agency in Frascati (RM, Italy). The authors want to thank Rosa Lasaponara from the Institute of Methodologies for Environmental Analysis, CNR-IMAA (C. da S. Loya, Tito Scalo, PZ, Italy), for her support and her interest in our research.

Buringh P. 1960. Soils and Conditions in Iraq. Baghdad.
Dore N, Patruno J, Sarti F, Ruescas AB, Hernandez M. 2010. Monitoring of UNESCO Sites in Danger, Including Samarra (Iraq) and Chan Chan (Peru), Using Optical and Radar Data. In Proceedings of EARSeL Symposium 2010, Remote Sensing for Science, Education, and Natural and Cultural Heritage, UNESCO, Paris (France), Reuter R (ed). University of Oldenburg: Germany.
ESA. 2009. EOLi ‘ESA’s Link to Earth Observation’. (accessed June 2009).
ESA. 2011. PolSARpro, the Polarimetric SAR Data and Educational Tool. (accessed December 2011).
Freeman A. 2007. Fitting a two-component scattering model to polarimetric SAR data from forests. IEEE Transactions on Geoscience and Remote Sensing 45(8): 2583–2592.
Web of Science® Times Cited: 27,
Freeman A, Durden S. 1998. A three-component scattering model for polarimetric SAR data. IEEE Transactions on Geoscience and Remote Sensing 36(3): 963–973.
Web of Science® Times Cited: 366,
Kendall TK. 2002. Napatan Temples: a Case Study from Djebel Barkal. The Mythological Nubian Origin of Egyptian Kingship and the Formation of the Napatan State. In 10th International Conference of Nubian Studies; Rome.
Lasaponara R, Masini N (eds). 2012. Satellite Remote Sensing: a New Tool for Archaeology. Series: Remote Sensing and Digital Image Processing, Vol. 16. Springer-Verlag: Berlin.
Lee JS, Pottier E. 2009. Polarimetric Radar Imaging. From Basics to Applications. CRC Press: Boca Raton.
Web of Science®
Nghiem SV, Yueh SH, Kwok R, Li FK. 1992. Symmetry properties in polarimetric remote sensing. Radio Science 27: 693–711.
Direct Link:
Web of Science® Times Cited: 89
Northedge A. 2007. The Historical Topography of Samarra. Samarra Studies. The British School of Archaeology in Iraq: London.
Patruno J, Dore N, Crespi M, Pottier E. 2013. Polarimetric multifrequency and multi-incidence SAR sensor analysis for archaeological purposes. Archaeological Prospection 20(2): 89–96. DOI: 10.1002/arp.1448.
Direct Link:
Full Article (HTML)
Patruno J, Dore N, Ruescas AB, Lavalle M, Putignano C, Jaia AM, Galdieri G, Loret E, Sarti F. 2009. Preliminary results of Alos Palsar imaging for investigation of archaeological underground marks. In Proceedings PolInSAR 2009, Workshop on Science and Applications of SAR Polarimetry and Polarimetric Interferometry. Frascati, Italy.
Piccarreta F, Ceraudo G. 2000. Manuale di aerotopografia archeologica. Metodologia, tecniche e applicazioni. EDIPUGLIA: Bari.
UNESCO. 1992. World Heritage List. (accessed October 2010).
UNESCO. 2003. Djebel Barkal and the Sites of the Napata Region, Description. (accessed December 2012).
UNESCO. 2006. Republic of Iraq Samarra Archaeological city, Nomination of Samarra Archaeological City for inscription on the World Heritage List. (accessed June 2012).
UNESCO. 2007. Samarra Archaeological City Documentation. (accessed December 2012).
UNESCO. 2011. Djebel Barkal and the Sites of the Napatan Region, Documents. (accessed December 2012).