Todd Brenningmeyer, Frederick Cooper,and Caitlin Downey
University of Minnesota
Todd Brenningmeyer is a graduate student in the Department of Art History, Frederick Cooper is a professor in the Department of Classical and Near Eastern Archaeology, and Caitlin Downey is a graduate student in the Department of Classical and Near Eastern
Archaeology at the University of Minnesota (Minneapolis).

Since 1990, the Minnesota Archaeological Researches in the Western Peloponnese (MARWP) project has undertaken two separate, concurrent projects around the peninsula that forms the southern half of mainland Greece. These projects explore a period that spans more than 3,000 years; participants seek to study and interpret a variety of architectural structures and features of the Peloponnesian landscape. The Morea Project consists of a survey of vernacular architecture in an area of the western Peloponnese -- known in the Middle Ages as the Morea -- from the period of Frankish occupation in the 13th century A.D. to the present. This project's goal is to discover, document, and study a variety of previously unrecorded buildings and towns before they are destroyed by man and nature. The Pylos Project involves the Bronze Age site of Pylos (circa 2,000­1,200 B.C.). This project, although concentrating on the reexcavation and production of a state plan of the palace complex, also includes a topographical survey of the ridge on which the palace was built. This article examines how applying and integrating GIS and Global Positioning System (GPS) technology is aiding the MARWP team in identifying undiscovered medieval sites and in interpreting unexcavated areas at the Pylos site.

GLOSSARY cbishoprics: an area over which a bishop has jurisdiction, similar to a diocese clintel: a horizontal, usually load-bearing, architectural member above an opening cmarly limestone: a crumbly limestone having similar properties to marl, a mixture of clay, sand, limestone, and often seashells cphytoarchaeology: a method of archaeology that seeks to define the relationships between particular types of vegetation and archaeological remains cscatterplot: diagram that plots out the density of specific spectral signatures ctholos: a beehive-shaped tomb having a horizontal passage, located on a hillside

MOREAN BACKGROUND

At the beginning of the 13th century A.D., large parts of Greece were conquered and settled by Crusaders returning to the west following the last crusades in the Holy Land. These conquerors, commonly called Frangkoi or Franks by the Greeks, were a mixed population of different European nationalities. They brought with them their own form of government and soon divided the conquered areas of Greece into feudal holdings and bishoprics. Although much of this conquered area soon regained self-governance, the Morea was occupied intermittently from 1205 through the 1440s. During the two centuries of Frankish presence, the conquerors constructed numerous churches, large fortifications, and domestic architecture in unfortified towns.

Figure 1

The Frankish occupation of the Morea had attracted little interest through the 1980s. As with other historical Greek periods, some Frankish remains have always been visible (see Figure 1). Written sources documenting this period of Frankish domination also exist, but many of these reside in archives in Venice, Italy, and Istanbul, Turkey, and are just now becoming accessible to interested scholars.

Earlier expeditions in the Morea occurring before MARWP had diverse goals and methods, and met with varying degrees of success. In the 1950s, scholar­cartographer Antoine Bon traveled the Morea and, with the guidance of medieval records, located a number of Frankish citadels and settlements that he plotted on a map of the region. He could not, however, locate many of the sites named in the records and thus only approximated their positions on his map.The locations of many of these sites remained unknown until the early 1990s.

Traditional pedestrian reconnaissance has also been used as a means of site recovery in the Morea. The Minnesota Messenia Expedition (MME) canvassed the southern region of the Morea in this way during the 1950s and 1960s, but that group's interests were focused on much earlier periods, particularly prehistoric. Other similar surveys have often overlooked sites because of limited ground visibility and the overall inaccessibility of site locations, many of which were built high in the mountains.

The majority of Frankish sites, however, have been located accidentally through excavation. Often, while searching for earlier Greek and Roman ruins, archaeologists discover Frankish remains. For example, during excavations at the ancient city of Corinth by the American School of Archaeology (Athens, Greece), trenches exposed Frankish buildings rather than the expected perimeter of a Roman temple and sanctuary. The Frankish occupation of Corinth dates from the 13th to the early 14th centuries A.D. and was undocumented until its accidental discovery in the 1980s.

Until a decade ago, architecture such as that represented by the Corinth site was essentially ignored by modern scholars. Scholarship focused primarily on the handful of previously documented sites located in areas of easy accessibility. The resulting effect was a skewed understanding of the Frankish presence in the Morea that favored sites that were easily accessible and previously documented. By the late 1980s the need for a more comprehensive study of the region was understood and MARWP began a program of site location and analysis.

SPATIAL INFORMATION A NECESSITY

Figure 2

Detecting and analyzing kastra (citadels) and accompanying towns occupying a geographic area of 10,000 square kilometers is an undertaking that simply cannot be accomplished with the methods used by Bon, the MME survey, or on-going excavation at a classical site such as Corinth. Integrating GPS technology with spectral analysis of remotely sensed data offers a superb alternative. Our study has adopted this methodology for archaeological prospection and analysis to cover a broad range of temporal and spatial data over a large and topographically diverse area. Obstacles such as hilly terrain, dense vegetation, and the inevitable destruction of built forms by both natural and man-made forces make it difficult to achieve a satisfactory level of comprehensiveness using traditional engineering survey techniques (see Figure 2 ). Only by fully integrating Landsat satellite image processing, GPS, and GIS technologies into our research program could we undertake these broad and varied projects.

We began our study in 1990 by consulting census records dating from the 1400s through the 1700s. Sites mentioned in these records, as well as others identified in Antoine Bon's study of the Frankish architecture of the Morea, were ground-truthed and sampled using GPS the following summer.

Behind the bushes. Methodologies developed in the field of phytoarchaeology were integrated into our research strategy to define relationships between particular types of vegetation and extant architectural remains. This relationship is expressed by a direct correlation between the abundant growth of prickly oak and holly near fortification walls. The examination of histograms and scatterplots of the near- and mid-infrared channels of the satellite scene has shown that the spectral response of the incumbent prickly oak, which obscures the walls from view, has a distinctive spectral signature that can be used to identify the presence of walls.

Figure 3 Figure 4 Figure 5

This correlation enabled the extraction of a reliable spectral signature for stone fortifications and citadel walls by image processing of our satellite scene in 1991. The location for Figures 3 and 4 is the Frankish site near Phigalea. In the aerial photograph of Phigalea (see Figure 3), the architecture has been obscuredby the vegetation; this same vegetation creates a spectral signature that allows us to detect the existing architecture below (see Figure 4). We then used this extracted signature to classify a subset of our satellite scene. In summer 1992, we tested the accuracy of our classification by ground truthing for sites identified in the lab. Because reliable topographic maps for the Peloponnese were not yet available, site location was limited to areas near recognizable topographic features. This early classification resulted in the location of two uncharted Frankish citadels in an area at the southwest end of the Erymanthos ridge, in the modern province of Achaia -- the citadels of Haghia Triada and Kastro Tis Oreias. On our arrival at Haghia Triada, we discovered structures that turned out to be fortification walls built into earlier walls of Hellenistic date (circa 400­200 century B.C.; see Figure 5). Other extant architectural remains include a massive rubble and mortar construction -- probably a keep -- and Frankish roof tiles, similar to others recently excavated from the Frankish site at Corinth. SETTLEMENTS AND SATELLITES Kastro Tis Oreias, the second site MARWP detected, rises above the gorge of the upper Peneios River. Like the citadel at Haghia Triada, it was an uncharted

site of some importance. The site is now desolate, occupied only by a single shepherd and his family, but seems to have been a thriving industrial settlement, as indicated by the remains of 240 houses, three churches, a cemetery, 16 mills along the river, and copious iron slag.

After this initial test application, we postponed further prospecting for new sites until 1996, at which time we continued testing our methodology and successfully located five possible sites. After verification through field reconnaissance, these sites were sampled and added to our base of recorded sites to be used for a supervised classification of our Landsat satellite scene. By incorporating a Trimble Navigation (Sunnyvale, California) base stationinto our sampling technique at this time, wedifferentially corrected sampled sites with an accuracy of 3­5 meters.

During the same season, one staff member meticulously scanned newly available Greek Army Geodetic Survey maps for suggestive toponyms located in terrain representative of possible Frankish occupation. The increased accuracy of these new topographic maps for the Peloponnese enabled us to georeference our satellite scene in the spring of 1997. We also extracted more accurate spectral signatures from our training sites to be used in our supervised classification.

Figure 6   Figure 7

Using universal transverse mercator (UTM) coordinates obtained during mapping, we compiled spectral signatures for specific architectural and archaeological features that were identified and sampled in the field. We used evaluation processes to obtain a tight overlap of spectral characteristics for highlighting those sets of pixels that conformed to the spectral signature of these training samples -- in other words, potential sites. This supervised classification has allowed us to define specific groups of pixels that conform to the outline of fortification walls and citadels (see Figure 6). We are currently in the process of fine-tuning our classification procedure by applying various spectral enhancements to our satellite scene of the western Peloponnese. We are currently using sites located through aerial photography and site toponyms as test sites for our classification. In August 1997, we ground-truthed an area of our classification that contained a high concentration of classified pixels. This area was not associated with a site toponym but appeared to have a strong likelihood for site location based on the clustering of "hot" or classified pixels. Using UTM coordinates gathered in the lab, a MARWP field crew successfully navigated to the location and verified the presence of a previously undocumented medieval settlement located near the small village of Ano Salmenikon (see Figure 7).

Our current prospection of sites through the analysis of remotely sensed multispectral images also allows a more comprehensive understanding of the macrogeographic relationships of medieval sites in the western Peloponnese area. Documenting previously unrecorded sites will make it possible to appreciate the principles that governed the development of self-protective networks over a large spatial and temporal range.

Along these lines, the integration of GIS coverage data showing locations of verified sites with an accurate digital elevation model (DEM) for the western Peloponnese will be used to determine spatial relationships involved in site placement. One such application is the development of line-of-site or viewshed maps to ascertain intervisibility relationships between sites within a given geographical area. We will use this method of analysis to test whether patterns of site location can be associated with particular types of architectural features with similar viewshed relationships. This type of macrogeographical analysis requires the availability of a comprehensive base of accurately mapped sites that has only been made possible through the image processing of remotely sensed multispectral satellite data and GPS.

MORE MOREA

Concurrent with our study of medieval Frankish citadels is our survey of standing vernacular architecture in the Morea. This project examines the methods and materials of construction that date from the Middle Ages, if not earlier, and which continued until the 1950s, when the widespread adoption of the concrete frame replaced these centuries-old construction techniques. We hope to develop a chronology of traditional Greek building as well as illustrate some of the methods by which architectural styles were transmitted across geographic areas.

To streamline the recording process over this large area, we have developed a method that combines traditional and electronic means of surveying and mapping. Using standardized survey sheets along with the GPS makes it possible to survey and record architectural features in the villages at a rapid pace.

Figure 8

Before the drawing team surveys a village, two to three persons walk through the village, marking a series of waypoints that are entered into the GPS as they go. This information is downloaded into AutoCAD (Autodesk, San Rafael, California) software and a preliminary map showing an array of waypoints is generated. Some of these points are joined to produce roads; others are interpolated to generate contours and topographic features (see Figure 8 right). The drawing team then uses these to navigate through each village as they record the buildings on the survey sheets. Team members sketch the plan and four facades of each house on these forms, noting the size, placement, and orientation of windows, doors, and outbuildings.

Figure 9 Figure 10 Figure 11

The encoded data collected for each house is then transferred to a combined graphics and database system, from which representations of each facade are produced. We are currently integrating this data into a GIS database. Using GIS we can initiate a query on individual architectural features -- such as the lintel in Figure 9 (left) or the architectural decoration in Figure 10 (left) -- that are found in a particular village, and relate them to any village in the database. For example, a certain style of lintel construction, dated to a particular period, can be queried. Any house that possesses this feature will be related to the query and the associated data can be displayed or printed for further analysis. The distance between houses with similar attributes is automatically calculated and displayed. This application of GIS technology has significant merit in the historical analysis of traditional vernacular architecture. We can chronicle the spatial distribution of a particular architectural style from one village to another, as well as analyze the interrelationships of this style with others that date to the same period and occur in the same villages. Using date stones such as the one in Figure 11, we can establish a chronology for specific architectural styles that can be input into our GIS database. The movement of individual stone masons and architects can thereby be associated with the movements and perceived economic viability of their contemporaries. Data integration. An additional application of this GIS resource is the integration of all of our current data sets into a single, unified whole. Coverage data illustrating the geographic locations of individual villages and house point locations can be integrated into our study of medieval Frankish settlements. An interesting application of this integration will be to chronicle how the development

of later town placement was influenced by the location of earlier Frankish settlements. In this way, the interrelationships that are detected between geographically removed areas of Frankish domination can be compared with those of later periods. The movement of trade networks, economic centers, and architectural innovations between areas of differing geographical and temporal range can thereby be related, allowing inferences to be drawn where congruous results are detected.

PALACE AMONG THE OLIVE GROVES

Figure 12

In addition to the Morea project, a MARWP crew has been working since 1990 at the Palace of Nestor at Pylos (see Figure 12), where we have integrated both GPS and GIS technology into our study of this Greek Bronze Age site. The Palace at Pylos, which was destroyed sometime near the end of the 13th century B.C., is the best preserved Bronze Age complex on mainland Greece.

The site itself was discovered in 1939 by Carl Blegen of the University of Cincinnati (Ohio) and was excavated from 1952 to 1966. Despite the length and intensity of these excavations, Blegen never made a detailed state plan of the site, nor a topographic map of the Ano Englianos ridge where the palace was built. In 1990 a MARWP team undertook the creation of a state plan of the palace and a topographic map of the ridge.

We began our survey of the ridge with traditional surveying techniques and tools. This process was time consuming; the landscape of the ridge also made the task practically impossible. Although the very top of the ridge, where the palace sits, was cleared for excavation, the rest of the ridge is covered with dense olive groves that obscure the line of sight much of the time. In following years, with the use of GPS, we created a topographic map of the ridge and site having an accuracy of 3­5 meters.

Figure 13

With this topographic map of the ridge (see Figure 13), we hope to use GIS technology to further interpret the site. We are currently pursuing a number of avenues with this. The first is a search for walls around the palace in areas that have not yet been excavated. The palace at Pylos is unique in that, unlike other contemporary structures, it does not appear to be enclosed by fortification walls. We hope to identify the remains of walls lying along the terraces that surround the palace, particularly those to the north, by overlaying Landsat imagery over our map of the Ano Englianos ridge.

Springs to come. Another area of interest will be the documentation of naturally occurring springs in the region. Several archaeologists and scholars who study Bronze Age religion have suggested that such springs may have carried a special religious significance during the Bronze Age period. In particular, springs have been discussed as possible locations for open-air sanctuaries. An understanding of their spatial distribution is therefore important, because detecting particular patterns of land use associated with specific cultural activities may be possible.

The GPS will be extremely useful in this survey. Waypoints generated near known spring locations may be downloaded into AutoCAD and then integrated into our GIS software packages -- AutoCivil (Research Engineers, Inc. [REI], Yorba Linda, California), ARC/INFO (ESRI, Redlands, California), and ArcView (ESRI) -- to create overlays of known springs. These overlays will be used with our Landsat satellite scene to provide a base for further prospecting by satellite remote sensing. Known springs, sampled using the GPS, will serve as training sites for a supervised classification of the area. Potential springs will then be ground-truthed for accuracy and added to our GIS database following their verification in the field. These data can then be integrated with our list of known structures and cultural features to determine how these areas were used in antiquity.

TALES FROM THE TOMBS

The third application of integrated GPS­GIS involves the chamber tombs associated with the palace. Although other spatial and social determinants are involved, the one thing common to many Bronze Age tombs in the southwest Peloponnese/Messene area is that the chambers were cut into outcrops of marlylimestone in a geophysical environment that is essentially marl and silt.

Figure 14

In cooperation with the Pylos Regional Archaeological Project (PRAP) associated archaeological survey team, headed by Jack Davis of the University of Cincinnati, we re-located the known tombs by UTM coordinates. In the fall of 1992, we created a GIS overlay of these sample tombs using the UTM coordinates collected with the GPS. A supervised classification of the sample tombs identified other possible tomb sites (see Figure 14). Again, developing an accurate signature for tombs is possible because of the unique

spectral responses of vegetation growing in a concentrated limestone environment. In future seasons and in association with PRAP, we may request permission to navigate additional areas of limestone outcroppings to verify sites identified in the lab and to search for other, undiscovered tombs. If this proves successful, we hope to identify a spectral signature for tholos tombs and will attempt to identify undiscovered tholos tombs using GIS overlays.

The development and analysis of data using techniques discussed in this article is an ongoing process that is currently in its infancy, both for MARWP as well as within the discipline of art history and archaeology as a whole. But even at this level, integrating GIS and GPS technology into our methodology has allowed us to expand beyond the limits and constraints of more traditional surveying techniques and has enabled us to interpret and expand far beyond the production of topographic maps.