You are here:Home/Research/ Byzantine Studies/ Project Grants/ Results of the Tophane Area GPR Surveys, Bursa, Turkey

Results of the Tophane Area GPR Surveys, Bursa, Turkey

Suna Çağaptay-Arıkan, University of Illinois at Urbana-Champaign / Bahçeşehir University, Istanbul, Turkey; Lawrence Conyers; and April Kamp-Whittaker, MA, Project Grant 2009–2010

See also Suna Çağaptay, “Results of the Tophane Area GPR Surveys, Bursa, Turkey,” Dumbarton Oaks Papers 68 (2014): 387–404.

This survey was made possible by grants from the Barakat Foundation at the University of Oxford, Dumbarton Oaks, and the Turkish Cultural Foundation. For all of these grants the authors are grateful.

The near-surface geophysical method called ground-penetrating radar (GPR) was chosen for an examination of buried architecture in the Tophane Park and surrounding areas of the Bursa old city. GPR is a non-invasive geophysical technique that can produce profiles and maps of buried features and stratigraphic layers. These are important considerations when working in complex urban sites like the Tophane area because of the complexity of ancient and recent building activity and renovations, many of which have been occurring for centuries.

The goal of the project was to identify areas that may contain Byzantine structures or their remains. The extant Byzantine structures in Tophane Park, the tombs of Osman and Orhan were destroyed in an earthquake in 1850 and subsequently rebuilt. They were reconstructed in approximately the same location and these buildings are visible today. This project's goal was to search for other remains of the original tomb and associated Byzantine structures nearby under the modern surface of the park, in an adjacent school playground, a tea garden, and a military complex. GPR has the ability to determine the depth and nature of stratigraphic layers, and can also determine what might be modern and what is more ancient.

Previous GPR Work

In 2007 a geophysical survey was conducted by Professor Metin İlkışık, working with the Municipality of Osmangazi. This survey used a 250 MHz antenna and sought deeply buried features and geological layers. The radar frequencies used in this survey penetrated to about 5.5 m. A cursory analysis of those data show that at about 3.5–5.5 m reflections were produced from a bedrock surface. The conclusion of this project was that the older remains rested on top of this bedrock layer and could be found at a depth of between 0.7–3.5 m. The 2007 survey also identified between 1.5–2.0 m of fill, which they associated with recent construction, and determined that the ancient remains begin at 1.5 m and continue down to the bedrock. They also noted that cables, pipes, and other modern utilities were predominantly located between 20–70 cm depth. This GPR survey supports that conclusion, and we found many modern intrusions such as cable and pipes but also important structural remains at a higher level in what was termed the fill layer for recent construction.

GPR analysis

Ground conditions in the tests varied from paved asphalt surfaces, cobblestone pathways, flagstones, marble flooring, and maintained grassy areas. Grids of GPR reflection data, composed of many reflection profiles were collected at all test areas. By collecting multiple closely spaced profiles, the reflection data can be interpreted both horizontally and vertically. In most cases horizontal amplitude-slices and vertical reflection profiles were produced from these profiles and both methods of viewing the ground were used in the interpretation. Nineteen grids of data were collected from June 7–23, 2009. A summary of those grids of data are shown below in Table 1. These grids were all chosen by Suna Çağaptay in areas of potential interest. Several were located in the immediate vicinity of the modern tombs and others scattered around the park. An additional series of grids were laid in the military zone and individual grids in the school yard adjacent to Tophane Park and also in the tea garden where salvage archaeological excavations had already occurred. Each data set is interpreted below in the order the GPR data were acquired.

GPR profiles and grids collected in 2009.
Grid Date of Collection Grid Description Number of Profiles Maximum Dimensions Profile Separation (meters)
3 6/9/2009 Grassy area in NW corner of Tophane park along military wall 33 16×10 0.5
4 6/10/2009 Grassy area in NW corner of Tophane park directly E of grid 3 32 16×18 0.5
5 6/11/2009 The interior of Orhan's Tomb 30 14×14 0.5
6 6/12/2009 This is composed of 4 separate areas, which are the paths around Orhan's Tomb. 53 34×33 0.5
7 6/12/2009 School yard 35 17×23 0.5
8 6/13/2009 This is a recollection of a section of grid 4. 22 11×9 0.5
9 6/15/2009 Flagstone surface in front of military office: Building #028 12 5×23 0.5
10 6/15/2009 Asphalt driveway leading to the military office 18 10×18 0.5
11 6/15/2009 Asphalt road and parking area to the South of the Military office 25 12×40 0.5
12 6/15/2009 Grassy area in the NW corner of the Military zone by the Byzantine wall 20 10×10 0.5
13 6/15/2009 Parking lot E of grassy area in Military zone 19 35×10 0.5
14 6/16/2009 Asphalt parking lot in front of main military building #004 51 43×25 0.5
15 6/16/2009 Residential parking lot on east side of military zone 21 11×50 0.5
16 6/17/2009 Military zone tea garden entryway 29 14×30 0.5
17 6/18/2009 Tea garden central pathway 37 61×18 0.5
18 6/22/2009 Dead end road behind Osman's tomb 6 33×3 0.5
19 6/22/2009 Cobblestone courtyard North of Orhan's Tomb 23 11×27 0.5

Table 1: GPR profiles and grids collected in 2009

Grids 1 and 2 are not listed in Table 1. These grids were two of the paths around Orhan's Tomb, and were recollected as part of Grid 6.

The GPR Method

The GPR measures the elapsed time between when pulses of radar energy are transmitted from a surface antenna, reflected from buried discontinuities, and then received back at another surface antenna (Conyers 2004). When the paired antennas are moved along transects on the ground surface, two-dimensional profiles of buried stratigraphy can be produced by stacking many hundreds or thousands of reflections together to produce what are termed reflection profiles. Changes in the reflected wave strength (measured as amplitude variations) and the geometry of those reflections in profile can then be related to the distribution and orientation of subsurface units and features of interest. These changes might be caused by stratigraphic layering, archaeological materials, anthropogenic soils or fill layers, and a variety of other objects or biogenic disturbances in the ground (Conyers 2004). Many tens or sometimes hundreds of reflection profiles, collected in a grid can then be analyzed within a three dimensional "cube" of reflection data as a way to produce complex images of buried materials (Conyers 2004, 148) in ways not possible using other near-surface geophysical methods (Johnson 2006).

Ground-penetrating radar is a geophysical technique that is most effective at buried sites where artifacts and features of interest are located within about 3 meters of the surface, but it has occasionally been used for more deeply buried deposits (Conyers 2004, 16). This depth of penetration and high degree of subsurface resolution make it a geophysical method particularly applicable to urban Turkey. Turkey's long history of habitation means that archaeological features can be both deeply buried and highly stratified.

A growing community of archaeologists has been routinely incorporating GPR as well as other near-surface geophysical methods for many years (Conyers 2004; Gaffney and Gater 2003). When this is done, GPR maps and images become primary data that can be used to guide the placement of excavations, or possibly to define sensitive areas containing cultural remains to preserve. Archaeological geophysicists have also used the GPR method to place archaeological sites within a broader environmental context, to test working hypotheses regarding past cultures, and to study human interaction with, and adaptation to, ancient landscapes (Conyers and Osburn 2006; Kvamme 2003).

The success of GPR surveys is to a great extent dependent on soil and sediment mineralogy, clay content, ground moisture, depth of burial, surface topography, and vegetation. It is not a geophysical method that can be immediately applied to any subsurface problem, although with thoughtful modifications in acquisition and data processing, GPR methods can be adapted to diverse site conditions.

Factors that affect GPR success

The resolution of buried materials and the depth of investigation are the most important factors that must be taken into account at all archaeological sites where the GPR method is contemplated. These two variables are inversely related and an analysis of them is crucial when choosing the appropriate frequency antenna to use for data collection. Higher frequency antennas, above about 400 megahertz (MHz) are capable of better subsurface resolution, but transmit energy to shallower depths (Conyers 2004, 39).

For instance, a 400 MHz antenna can resolve objects and stratigraphic interfaces as small as about 20 centimeters in maximum dimension, but are only rarely effective below depths of 3 meters. In contrast, lower frequency antennas (in the 100–200 MHz range) can theoretically transmit energy that penetrates 5 meters or more, but are incapable of resolving objects or interfaces smaller than about 60 centimeters in dimension. In most of the ground conditions in the Tophane area there was good depth penetration to about 1.5 m or even greater using the 400 MHz antennas. Below that depth radar energy was attenuated by electrically conductive materials in the ground and obscured somewhat by ambient electromagnetic noise (cell phones and other radio broadcasts) created by the urban environment.

Radar energy attenuation with depth is mostly a function of the electrical conductivity of soils and sediments through which the radar energy must pass (Doolittle and Collins 1995). Highly electrically-conductive material effectively destroys the transmitted radar energy at shallow depths by removing the electrical component of the electromagnetic wave, causing propagation to cease (Conyers 2004, 49). Radar energy loss, termed attenuation, always occurs as energy moves into the ground. This attenuation is a function of four general factors (Reynolds 1997). Coupling losses occur when the radar antennas are not placed in direct contact with the ground, or when the ground surface is uneven, allowing radar energy to be scattered and lost before it effectively "couples" with the ground to be transmitted within it. This loss factor can be mostly overcome by making sure antennas are moved slowly and carefully along the ground surface. The surfaces in the Tophane area survey were well-suited for GPR survey and coupling was not an issue.

Another factor is the geometric spreading that occurs as energy moves into the ground. This loss is a function of the conical shape of the transmitted radar pattern that spreads the energy out over a larger and larger surface area as it travels deeper in the ground (Conyers 2004, 62). Spherical spreading with depth decreases the amount of energy that can be reflected back to the surface from any one buried object or interface below the surface, lowering the effective resolution of any reflections generated from it. This is a factor inherent in the method and cannot be adjusted for using standard GPR equipment.

A third site-specific factor is energy objects or discontinuities in the ground, redirecting some of the energy away from the surface receiving antenna so that it is not recorded. This was not a problem in the Tophane area as the ground tested was not composed of large-sized stones. A similar site specific factor, and the one that is the most variable and important factor in determining the GPR method's effectiveness, is electromagnetic attenuation. As radar energy is composed of both electrical and magnetic waves, which move in a conjoined fashion (Conyers 2004, 24), the removal of either one or the other by electrically conductive or magnetically susceptible ground effectively destroys the transmitted energy. In general, soils that are moist and have high clay content, especially clays of certain mineralogy, will have high electrical conductivities as measured by their cation exchange capacity (CEC). While no chemical tests were performed on the soils and sediments in the Tophane area, visual analysis showed them to be composed of sandy silts with some clay. The question of the soil composition of the Tophane area could be further resolved through an examination of geological studies of the area. This could provide more information of the attenuation of the radar frequencies at 1.5 m depth. For the most part, however, penetration depth in the Tophane area was sufficient to resolve the features of interest, and energy attenuation was not a negative factor.

Tophane area GPR Tests

In all tests the Geophysical Survey Systems, Inc. (GSSI) Subsurface Interface Radar System model 3000 (SIR-3000) was used to collect the GPR data, with a survey wheel used to place reflections in space along survey transects. The 400 MHz antennas were used in all tests. Reflection data were transferred to a laptop computer and processed using software that is publicly available (Conyers 2005). This software allowed reflection profiles to be viewed and analyzed for effective depth penetration, and at some sites grids of closely spaced profiles were used to produce amplitude maps of buried features of interest.

Data processing procedures

The raw reflection data collected by GPR is nothing more than a collection of many individual traces along two-dimensional transects within a grid. Each reflection trace contains a series of waves that vary in amplitude depending on the amount and intensity of energy reflection that occurred at buried interfaces. When these traces are plotted sequentially in standard two-dimensional profiles, the specific amplitudes within individual traces that contain important reflection information are sometimes difficult to visualize and interpret. The standard interpretation of GPR data, which consists of viewing each profile and then mapping important reflections and other anomalies, is rarely sufficient, especially when the buried features and stratigraphy are complex. For this reason, amplitude analysis must be used.

An analysis of the spatial distribution of the amplitudes of reflected waves is important because it is an indicator of subsurface changes in lithology and other physical properties of the ground. The higher the contrasting velocity at a buried interface, the greater the amplitude of the reflected wave from a surface or point source object (Conyers 2004). If amplitude changes can be related to important buried features and stratigraphy, the location of higher or lower amplitudes at specific depths can be used to reconstruct the subsurface in three dimensions. Areas of low amplitude waves indicate uniform matrix material or weathered and homogenized A and B soil horizons, while those of high amplitude denote areas of high subsurface contrast such as buried archaeological features or possible floor features. In order to be correctly interpreted, amplitude differences must be analyzed in discrete slices that examine only the strength of reflections within specific depths in the ground. Each slice consists of the spatial distribution of all reflected wave amplitudes at various depths, which are indicative of these changes in sediments, soils, and buried materials.

Amplitude slices were constructed parallel to the ground surface in equal time intervals of 5 nanoseconds (TWTT). Each 5 ns slice then became analogous to mapping about 30 cm in thickness, similar to arbitrary elevation levels in standard archaeological excavations. In the analysis of the slice maps from the Tophane project, I did not analyze the top slice maps because these maps contained primarily modern disturbances such as pipes and the historic features of interest were not distinguishable. I was also working on the assumption that Byzantine features would not be found in the top 30–60 cm of soil.

To compute horizontal amplitude slices the computer compares amplitude variations within traces that were recorded within defined time windows. When this is done both positive and negative amplitudes of reflections are compared to the norm of all amplitudes within those windows. No differentiation is usually made between positive or negative amplitudes in these analyses, but only the magnitude of amplitude deviation from the norm. Low amplitude variations within any one slice denote little subsurface reflection and therefore indicate the presence of fairly homogeneous material that produced few if any reflections. High amplitudes indicate significant subsurface discontinuities, in many cases detecting the presence of buried features that produced strong reflections at the boundary between the two (in this case stones or possible floors or walls). An abrupt change between an area of low and high amplitude can be very significant and may indicate the presence of a major buried interface between two media. Degrees of amplitude variation in each time-slice are assigned arbitrary colors along a scale. Usually there are no specific amplitude units assigned to these color or tonal changes. In all subsequent maps I have used a rainbow of colors in the amplitude maps, with red being highest amplitude (presence of rocks or hard surfaces) and blue and white being homogeneous soils. Many of the grids collected are not a uniform rectangular shape. When the computer processes these grids the areas of missing data are filled in by the program. These areas of the map appear as light blue or white in color and indicate the missing data.

When interesting buried high amplitude features show up in slice maps, I always go back and look at the individual profiles to confirm what these features might be, and also study what they look like in profile. When significant features are visible in both map form and profile, those will be shown as figures in the discussion of each grid of GPR data.

Results of the Surveys by Grid

Grid 3

Grid 3 was located in a grassy area in the northwest corner of Tophane Park along a large wall that separated the park from the adjacent military area. The area is moderately landscaped with small shrubs and trees. This area was chosen because it was thought to possibly contain part of the Byzantine inner fortifications.

Amplitude maps of Grid 3.
Fig. 1: Amplitude maps of Grid 3.

Amplitude slice maps show levels of 5 ns or 30 cm thickness in horizontal layers of the ground. Each depth of 30 cm is approximated using a relative dielectric permittivity of 6.3 shows a two-way travel velocity of 6 cm/nanosecond, derived from hyperbola fitting (Conyers 2004).

The amplitude slice maps of the grid revealed very little. Along the west side of the map running from y=0–12 there are a series of strong reflections created by architectural fill at the base of the existing wall. At y=6, x=2 the reflection of a possible wall segment is visible in the 17–22 ns slice as an area of high amplitude reflection. The high amplitude reflection running N/S at x=4 is a modern pipe line.

Grid 5

Grid 5 was collected in the interior of Orhan's Tomb. The current tomb was constructed after the earthquake destroyed the original structure. It is presumed that at least portions of the original tomb lie under the modern structure. In several places in the current building areas of older mosaic flooring is still visible. The center and eastern side of the structure is filled with raised sarcophagi and the grid was collected around these obstacles.

Grid 6

Grid 6 is actually composed of four different regions surrounding Orhan's Tomb. All four regions were collected using the same parameters so that the grids could be compared visually.

It was hoped that these grids would find remnants of the previous structures which were destroyed in the earthquake. The area has been extensively landscaped both during the reconstruction of the tombs and during subsequent remodeling.

Area 1 is the largest of the GPR grids and covers the east side of the tomb. This grid was composed of cobble stone pathways with raised plant beds in between. It was not possible to collect data in the plant beds; consequently, the grid has an irregular shape with data missing in the areas where the plant beds were located.

Grid 17: Tea Garden

Map of Grid 17 with salvage excavation walls. The red outline represents the shape of the grid, with the black dot as 0,0. The blue rectangles are the locations of walls visible in the salvage excavation and where they would intersect the GPR grid. Some of these walls were visible in the grid maps.
Fig. 2: Map of Grid 17 with salvage excavation walls. The red outline represents the shape of the grid, with the black dot as 0,0. The blue rectangles are the locations of walls visible in the salvage excavation and where they would intersect the GPR grid. Some of these walls were visible in the grid maps.
Amplitude map of grid 17 generated at 30 cm horizontal intervals.
Fig. 3: Amplitude map of grid 17 generated at 30 cm horizontal intervals.
This grid was collected down the center of a park area and tea garden several blocks east of the Tombs. A majority of the grid was collected along the sidewalks that run east through the park. There are two major sidewalks parallel to each other that run along either side of a central pool and grassy area. The pool was not collected but the grassy area between the sidewalks was. This created a U-shaped grid. The area was chosen because in 1986 and 2000 a series of salvage excavations was conducted. There are still exposed walls visible from these excavations that date either from the Byzantine or early Ottoman periods. The grid collected ran along these salvage excavations. Several of the walls visible in the excavation grids can be seen continuing into the GPR grid. The walls visible in the salvage excavation vary greatly in their depth below the surface. Some are almost directly under sidewalks and other modern features while some walls become visible at 1.5 m. This variability in depth is reflected in the amplitude maps. The level of variability also makes dating features difficult. It means that we can not establish a depth at which Byzantine features will be found below. The 10 ns approximation of the earthquake surface is still a good estimate to eliminate many of the most modern surface disturbances. However, this grid demonstrates the need for more specific data about the depth of different historic layer, especially from the Byzantine period, in order to conclusively date any of the features visible in the GPR grids.

Reflection profile 17, x = 8.5 in the amplitude maps, showing several of the features visible in the slice maps. At 2 m a wall is visible. This is feature #4 from the slice maps. Between 6–10 m there is a floor surface visible at the surface of the grid. This is a modern sidewalk. At 18 m there is a deeply buried wall at 30 ns. This is feature #6 from the slice maps. Finally at 20 m there is a floor surface visible that is part of feature #2, the surface visible in the 17–22 ns horizontal slice map map.
Fig. 4: Reflection profile 17, x = 8.5 in the amplitude maps, showing several of the features visible in the slice maps. At 2 m a wall is visible. This is feature #4 from the slice maps. Between 6–10 m there is a floor surface visible at the surface of the grid. This is a modern sidewalk. At 18 m there is a deeply buried wall at 30 ns. This is feature #6 from the slice maps. Finally at 20 m there is a floor surface visible that is part of feature #2, the surface visible in the 17–22 ns horizontal slice map map.
Some of the walls were visible in the grid maps.

  1. A small room visible in the 17–22 ns slice.
  2. A floor surface that stretches over a large area of the grid. There is also a water faucet at this point in the grid so this is making the feature appear more extensive. The depth of this feature is fairly shallow but in the reflection profiles the floor surface is clearly visible.
  3. A large E/W running wall which can be seen in the 17–27 ns amplitude maps.
  4. A smaller diagonal wall segment visible in the 22–27 ns maps.
  5. Two intersecting walls. One runs E/W and the other N/S. The N/S wall aligns with a wall visible in the salvage excavations.
  6. Another large wall running N/S. It can be clearly see in 32–42 ns maps.
  7. A minor reflection that appears to be the corner of a wall visible in the salvage excavations.

The circular open area visible in the first 3 maps at 10–15 m W is the location of a large tree in the grid.