Results and Interpretation

Results and Interpretation

The following geophysical datasets were collected in the study area: six hectares of GPR (Figure 5), 4 hectares of magnetometry (Figure 5), and 1.6 hectares of EMI (magnetic susceptibility and conductivity) (Figure 6). These data provide a vast database of information that cannot be completely described here, so we will focus on a few key geophysical features. Some of these features have been confirmed by excavations, which were placed to test GPR anomalies as part of the second author’s master’s thesis (Koons 2006). Other anomalies are described using quotations where necessary to signal that they have not yet been tested by excavation. The main geophysical features include circular water retention features that may have been “pools” or “baths”, a series of water conduits, “residential compounds”, small segments of wall foundations (probably razed remains from urban renewal), a gravel pavement, “revetments” aligned with the cardinal directions, and a “square structure.”

Figure 5. (a) Extent of GPR survey on the east side of Akapana, about six hectares; and (b) extend of magnetometry surveys, about four hectares.

Figure 6. Extent of (a) magnetic susceptibility and (b) conductivity surveys in the study area, about 1.6 hectares.

“Pools”

A series of circular anomalies located due east of the Akapana were detected with all geophysical data types, each providing valuable clues for interpretation. The largest of these is shown in Figure 7. GPR reflections around the perimeter suggest that the edges slope toward the center (Figure 7b). The dearth of reflections from inside the circle are probably due to the fact that this area has more clay than surrounding sediments, causing attenuation of the GPR signal and limited depth penetration. In addition, if the sloping edges continue toward the center, they may be beyond the depth capabilities of GPR in this area. Magnetometry data show several large dipolar anomalies on the outside of the circle, with only a few small, weak anomalies inside, suggesting that this feature is surrounded by blocks of andesite (Figure 7c). Conductivity data indicate that the interior, particularly the southwestern quadrant, contains more clay and/or moisture than surrounding materials (Figure 7d). Magnetic susceptibility is low inside the feature, suggesting that the topsoil has been removed or heavily disturbed (Figure 7e). Finally, a 1952 aerial photo shows vegetation growth concentrated around the perimeter. Together these data suggest that the feature is a circular or semi-circular water retention feature or “pool” surrounded by blocks of andesite.

Figure 7. Possible pool discovered east of Akapana: (a) location of the most prominent circular anomaly within the GPR survey area; (b) GPR data show a circular pattern with strong reflections around the perimeter and none on the inside; (c) magnetometry data suggest that this feature is surrounded by large blocks of andesite, with none inside the circle; (d) the interior of the circle is relatively highly conductive, suggesting moisture retention and/or clay deposits; (e) magnetic susceptibility is low inside the circle, suggesting that topsoil was removed; (f) 1952 aerial photo shows thick vegetation growing around the perimeter. All of this suggest that this is a water retention feature surrounded by blocks of andesite.

At first this interpretation seemed unlikely, but further research indicated that indeed this type of feature is known from Tiwanaku. German archaeologist Max Uhle took photos of two such features in 1893 (Figure 8). The first is a monumental spring located south of Pumapunku, where andesite blocks surround a depression created to collect water from a natural spring. The second is very similar, but the exact location at Tiwanaku is not known. These photographs show nearly an exact match to the information gathered with geophysics and strongly suggest that a number of these “pools” were built east of Akapana. Test excavations on the southern edge of this feature were placed to test the GPR data (Koons 2006) and ancountered an abundance of moist, clay-rich sediment. This further supports the interpretation. Future excavations utilizing the other geophysical data sets are needed, however.

Figure 8. Examples of “pools” at Tiwanaku known from photos by Max Uhle (1893). (left) located south of Pumapunku, this feature is composed of blocks of andesite surrounding a depression where spring water collects. (right) A similar feature found at Tiwanaku, exact location unknown. (photos courtesy Alexei Vranich).

“Residential Compounds”

Two possible residential compounds have been identified in the geophysical data. One is located east of the modern fence, southeast of the Kantatayita (Figure 9a). Subtle lineations are discernable in the GPR data (Figure 9b), but these are much clearer in magnetometry (Figure 9c). Together the two datasets suggest a rectangular compound measuring 33 m east-west and 27 m north-south. This agrees with projections by Williams et al. (2007), who found that a residential compound located east of the present geophysical survey area was about 30 meters wide. The magnetic anomalies are relatively weak and may indicate wall foundations built of moderately magnetic stone (perhaps sandstone) and/or baked adobe. The southern wall consists of parallel lineations, perhaps representing two separate courses of foundation stone, which is typical for residential compound walls elsewhere at Tiwanaku (Janusek 2003; Kolata 2003b). Magnetic anomalies inside the “compound walls” could represent burned features, such as cooking or firing areas, or large pieces of andesite.

Figure 9. “Residential Compound” interpreted based on GPR and magnetometry anomalies. (a) location of anomalies within GPR survey; (b) anomalies in GPR that suggest a rectangular feature (green arrows); (c) magnetometry data have anomalies in the same locations, plus a more complete rectangular outline with more detail of the west and south boundaries (pink arrows). The size and orientation of these anomalies suggest that this is a residential compound, and the southern wall appears to have two parallel components consistent with other residential compound walls at Tiwanaku. Interior magnetic dipolar anomalies could indicate cooking or firing areas, or blocks of andesite.

Wall Foundations

The GPR data show many short linear anomalies oriented in the cardinal directions. A handful of these were tested in 2005 (Koons 2006), revealing portions of wall foundations. Two excavation units were placed west of the Kantatayita (Figure 10a-c). Unit S2 revealed a wall foundation oriented north-south, consistent with the anomaly (Figure 10c). More wall foundations were found in Units S1 and S8 (Figure 11). In both cases GPR anomalies proved to be very reliable indicators of subsurface architecture. Not only were the wall foundations encountered exactly where predicted, but they disappear exactly where the GPR anomalies disappear.

Figure 10. Portions of wall foundations discovered by testing GPR anomalies. (a) location of magnified area, (b) magnified area showing the Kantatayita (where no GPR data were collected), the 40 x 40 m GPR survey region, and two excavation units measuring 5 x 5 and 5 x 7 meters; (c) magnified GPR slice showing anomalies that were tested and excavation units (after Koons 2006); (d) unit S2 showing a wall foundation indicated by the GPR anomaly (after Koons 2006).

Figure 11. More wall foundations encountered when testing GPR anomalies. (a) GPR slice showing anomalies tested by excavation units S1 and S8; (b) wall foundation encountered in these units, consistent with the GPR anomalies. (after Koons 2006)

Conduits

Additional linear GPR anomalies oriented obliquely were also tested with 5 x 5 m excavation units (Koons 2006) (Figure 12a-b). Both units encountered water conduits beneath a clean gravel pavement situated 45-50 cm below the surface (Figure 12c). As with the wall foundations discussed above, these GPR anomalies proved to be very reliable indicators of subsurface architecture. The linear anomalies representing wall foundations and conduits are very similar in the GPR data, but can be distinguished by their grid orientation. All of the wall foundations are oriented with the cardinal directions, but canals are oriented obliquely. This relationship may not be foolproof, but is one way to interpret anomalies such as those indicted in Figure 13.

Figure 12. Test of oblique linear GPR anomalies. (a) location of magnified area; (b) magnified GPR image showing locations of excavation units; (c) results of two 5 x 5 m test excavations revealing conduits as the source of the oblique linear GPR anomalies (after Koons 2006).

Figure 13. Linear GPR anomalies west of the Kantatayita oriented at various angles. Those oriented with the cardinal directions are interpreted as wall foundations or other components of architecture, while obliquely angled anomalies are more likely to represent conduits. (a) location of magnified area; (b) magnified portion of GPR slice composite showing linear anomalies.

Gravel Pavement

The gravel pavement has been found in all excavation units in this area (Cortez Ferrel 2006; Koons 2006). It is composed of gravel mixed with clay, which should cause high amplitude GPR reflections. Yet, the pavement is not visible in the GPR data in most locations for two reasons. One problem is contrast. Archaeological features are detected with geophysics because they contrast with immediately surrounding deposits. Since the gravel layer extends over a very large area (indeed it could be larger than the GPR survey area), there is no contrast from one place to the next. In other words, when something exists everywhere it becomes part of the background and will only be detected on its edges. Also if the pavement is detected it will create flat-lying reflections, which can easily be confused with noise and removed with a standard “background removal” filter (Conyers 2004). Figure 14 illustrates this problem. The upper profile (Figure 14a) contains horizontal banding from noise (unwanted radio signals recorded by the receiving antenna), but also contains reflections from horizontal layers in the ground. When a background removal filter is applied both the noise and flat-lying reflectors are removed (Figure 14b). The result is a general improvement compared to the original data but extensive flat-lying reflections are also removed. We are currently working on a new background removal filter that does not remove the gravel pavement reflections.

Figure 14: Accidental removal of horizontal reflections with background removal filter. (a) a reflection profile before background removal, showing horizontal banding related to noise, and a flat-lying reflection that could be archaeological interest (outlined in red); (b) after background removal the horizontal banding is removed, but so is the flat-lying reflection that is probably not related to noise.

Excavation unit S3 (Figure 12b) was placed to test a broad amorphous GPR anomaly, which appears in profile as a sloping reflection (Figure 15a). No corresponding sloping feature was encountered. Instead, the flat-lying gravel pavement was found throughout the unit (Figure 15b). The modern ground surface above, however, is gently sloped where the anomaly appears. Since the GPR data have not been topographically corrected, the reflection profiles treat all subsurface reflections as if they occurred beneath a perfectly flat surface. Since the ground is sloped in this area, a flat-lying reflection will appear sloped and will be the mirror-image of the ground above. Therefore, the southward-dipping reflection occurs where the surface slopes down northward, and the reflection actually represents the flat-lying gravel layer. Thus, the gravel layer becomes highly visible in GPR reflection profiles and slice maps wherever the surface is sloped (Figure 15c). The digital elevation model (DEM) created by Barnes and Cothren (2007) is not detailed enough to show this slope, but we can infer ground surface slope based on the GPR data and field notes taken during geophysical data collection and excavations. It is unclear at this time if the surface topography is related to the buried urban landscape or more recent erosional processes.

Figure 15. Gravel pavement layer visible where modern surface is sloped. (a) reflection profile showing a dipping reflection that corresponds to the horizontal gravel pavement (after Koons 2006); (b) 5 x 5 m excavation unit showing the flat gravel pavement and sloping ground surface, with left red arrow indicating location of profile in “a” (after Koons 2006); (c) portion of GPR survey area showing GPR anomalies created by the gravel pavement underneath the sloping modern surface. Green arrows point out the anomalies and indication the direction of surface slope (north and west).

“Revetments”

The northern portion of the survey area contains several very long linear anomalies in GPR (Figure 16a) and magnetometry (figure 16b), which are aligned with the cardinal directions. A 1952 aerial photo also shows vegetation marks that coincide with the geophysical anomalies (Figure 16c). Figure 16d is a composite of the two geophysical datasets, with magnetometry in the background and GPR superimposed on top and 50% transparent. Anomalies described in the text are labeled A-K on the appropriate images in Figures 16. Anomaly “A” is a revetment (retaining wall) built of a variety of stones including andesite, known from excavations by a French expedition in 1903 (Stanish 2002). This revetment probably was built to stabilize the edge of a terrace to the west. Part of the revetment (“B”) that was not excavated is clearly visible in GPR and the aerial photograph. These anomalies are somewhat complicated because a modern pedestrian pathway is located along the west side of the excavation trench (now filled), and in other locations in this area (“C”).

Figure 16. Northern portion of survey area showing possible revetments and other features. (a) ground penetrating radar; (b) magnetometry; (c) 1952 aerial photograph; and (d) combination of magnetometry (blue) and GPR (magenta). Key: A = 1903 excavation trench and revetment; B = unexcavated portion of revetment visible in GPR data; C = location of modern pedestrian pathways; D = possible revetment constructed with andesite or large water conduit; E = possible revetment; F-L = possible east-west revetments constructed of sedimentary rock or some other non-magnetic material.

Magnetometry data and the 1952 aerial photograph suggest a second revetment is located about 40 meters to the east (“D”). It is also visible in GPR, but is very difficult to see in the slice composite. This linear feature is interpreted as a revetment because it is similar and parallel to the known revetment to the west. Alternatively, it could be large, deep conduit like others found at Tiwanaku (Couture and Sampeck 2003). The strong magnetic dipoles suggest that this feature was built using large pieces of andesite. A third possible revetment (“E”) is indicated by GPR, magnetometry, and the 1952 aerial photograph. The GPR anomalies indicate that the surface slopes toward the east, perhaps indicating that the modern surface still preserves some of the ancient topography created by terrace and revetment construction. Magnetometry indicates that some portions of this feature may include andesite.

A series of east-west lineations (“F” – “L”) also occur in this area. These features are clearly shown in the GPR data, and some are apparent in the 1952 aerial photograph, but they do not show up in magnetometry suggesting an absence of andesite and other igneous rocks. They could be built of sandstone or some other weak- or non-magnetic material. These lineations could also be revetments, and they seem to make connections between the north-south “revetments.” Together these long linear anomalies suggest a series of terraces bounded by revetments descending east from the Semisubterranean temple and north from the Akapana. The broad area between “D” and “E” may indicate a plaza, as the DEM indicates that this area was relatively flat compared to the surroundings (see Barnes and Cothren 2007).

“Square Structure”

North of the east-west “revetments” is a possible square structure indicated by GPR (Figure 17). This geophysical feature is very difficult to interpret because there are few similar features in the archaeological record. It measures about 20 meters across and is roughly square, much like the Semisubterranean Temple to the southwest (Figure 1). An absence of magnetometry anomalies suggests that it does not include any andesite, but large dipolar anomalies to the west may indicate andesite blocks associated with this feature. The newly created DEM (Barnes and Cothren 2007) shows a slight topographic rise in this area, further suggesting the presence of buried architecture. Without many similarly sized and shaped features in the archaeological record, and without additional geophysical data to the north and east it is very difficult to interpret this feature. Test excavations are needed.

Figure 17. “Square structure” indicated by GPR. (a) northern portion of GPR survey area showing location of a possible square structure; (b) magnetometry data in same area does not show the square feature (indicated in red based on GPR), but anomalies to the west could be related.