The University of Bradford's Department of Archaeological Sciences provided us with a Geoscan FM-18 fluxgate gradiometer to gather geomagnetic data. The FM-18 is a sturdy machine, designed specifically for the rigors of archaeological field work. It employs a pair of fluxgate sensors to measure the vertical gradient intensity of the earth's magnetic field, eliminating the need to correct the data for magnetic inclination. The instrument's automatic calculation of the vertical gradient intensity (done by taking the difference between the two sensors' readings) also eliminates the need to correct the data for diurnal variation. The FM-18 also allows data to be zeroed to a reference point so that separate data sets can be readily joined. To gather data, one simply walks, holding the instrument as vertical as possible, and presses a button to log a data point into the instrument's memory. The instrument contains enough memory to store 3200 data points.
A regular pattern of data collection must be employed in order to produce a useful map of the geomagnetic data. The FM-18's recording device allowed several types of collection strategies. We chose to collect data in square units or grids. Each grid consisted of twenty evenly-spaced lines, each line containing twenty evenly-spaced data points. In this manner, the magnetic character of each grid is represented by four hundred evenly-spaced readings of the earth's magnetic field. Data were collected for two kinds of grids: (1) 20 by 20 m grids with lines and data points at 1 m intervals; and (2) 10 by 10 m grids with lines and data points at .5 m intervals. Corner points for each grid were established with a theodilite and electronic distance measurer (EDM) to accurately locate them on the site. The grid lines themselves were physically laid out on the ground surface using nylon ropes with . 5 m markings. All lines were oriented north-south, and all magnetic data were collected walking south to north along the lines, moving west to east as each line was completed.
In 1993, data collection took place over a one month period from May 20 to June 19. A total of sixteen days were spent collecting geomagnetic data during this time period. Data were gathered for 93 grids, a total of 37,200 data points, in seven main blocks [cross-reference TSW topo]. Block 1 covered the largest area, a total of thirty-six 20 by 20 m grids directly east of the inner town and including sections of both the inner and lower town wall (see discussion below). Block 2 was the smallest, a single 20 by 20 m grid between Blocks 1 and 3. Block 3 consisted of eight 20 by 20 m grids north of the inner town and crossing the lower town wall. Blocks 4, 5, 6, and 7 each consisted of four 20 by 20 m grids, with Blocks 4, 5, and 7 each sampling a different area of the lower town, and Block 6 sampling an area of the lower town's southern extension. One grid in each of these blocks was surveyed at both 1 m and .5 m intervals to determine the degree to which closer data point spacing increased resolution, and four grids in Block 3 were surveyed twice to provide data for establishing reliability of the survey.
In 1995, data collection took place over a three week period from June 5 to June 29. A total of eighteen days was spent collecting geomagnetic data during this time period. Data were collected for 114 grids, a total of 45,600 data points, in two major blocks [cross-reference TSW topo]. Block 8 covered an area of 200 by 160 m on the northwestern edge of the lower town. Block 9 covered an area of 160 by 80 m crossing the southern edge of the lower town and extending into the town's hypothesized southern extention. We chose to gather data over several large, contiguous blocks, rather than covering smaller, but more diverse areas of the site, because our 1993 work demonstrated that it was much easier to interpret larger areas than small ones. While this strategy limited the range of locations we could investigate, we felt it would allow us to better understand the data we did collect.
To image subsurface features, the geomagnetic data were downloaded from the gradiometer and plotted as dot-density and contour maps. The maps' x- and y-coordinates were established by the magnetic survey grid and linked to the Sweyhat site grid by recording the corners points of the magnetic survey blocks. The z-coordinate was the vertical gradient intensity of the earth's magnetic field at a particular grid point. In essence, the final product of the geomagnetic surveys was a set of contour maps of the vertical gradient intensity of the earth's magnetic field over particular areas of the site. We used a DOS-based laptop computer running the Geoplot software package to download the data and do initial field mapping (in this case, dot-density mapping). At the University of Pennsylvania Museum, we exported the Geoplot files to Microsoft Excel for Macintosh so they could be read into MacGRIDZO, a contouring application.
Each file output from Geoplot consisted of 400 readings of the intensity of the earth's magnetic field (one magnetic grid), which we transferred to Macintosh in ASCII format. We then assigned x- and y-coordinates to the z-values for plotting in MacGRIDZO. The readings for each grid had been calibrated to a zero point to correct for instrumental "drift" during data collection. These readings were output in the order collected, and took the form of a 400-line column of raw numbers. To prepare these for mapping we imported them into an Excel spreadsheet which had been set up with the correct x- and y-coordinates as the first two columns, and to which we simply added the 400-line column of raw data as a third column. This was made considerably easier by the fact that we had collected all the data in the same way (moving south to north and west to east). We also used Excel to concatenate magentic grid files into files representing our larger magnetic blocks before plotting them in MacGRIDZO.
We chose MacGRIDZO as our primary mapping package because of its choice of interpolation algorithms (inverse distance and moving weighted least squares) and its flexibility in data manipulation, particularly in terms of selecting data ranges for mapping and in adjusting input and output parameters, e.g., combining grid files and filtering data. Much of MacGRIDZO's flexibility comes from the way it handles data. It first generates a "grid file" by taking the raw data and dividing it up into a grid of imaginary cells. This "grid file" might be conceptualized as a mesh draped over the raw data file with z-coordinate values assigned to each individual grid node on the mesh via the user-specified interpolation algorithm.
The number of cells in a grid file (user-specified) determines the total number of intersections in the mesh (nodes), and thus the total number of interpolated values used in generating the contour map. The number of cells plays a major part in the resolution of the final image. We found that an optimal number of cells for most applications was 900: 30 cells along the x-axis and 30 along the y-axis. In larger files, like our Block 1, cell spacing was higher (fewer number of cells and thus lower resolution) owing to the size and processing time required to generate such a large file.
For interpolation of the data we chose the moving weighted least squares method, which works the same as the inverse distance method. The inverse distance method estimates a grid node's value using the weighted average of a user-specified number of neighboring data points, which are weighted using the inverse of their distance to the grid node, taken to a user-specified power. These values are then used to determine a first order polynomial (a floating plane) for each grid node. New grid node values are the computed based on the intersection of the node with its plane.
Although MacGRIDZO provides a variety of output options, we used it primarily to create two-dimensional color contour plots. MacGRIDZO generates color-filled contours depending on the user-specified contour interval entered before the image is produced. For example, a value of ten will produce a contour map with contour lines drawn for every ten z-units. We found a contour interval of 0.25 nT produced maps of sufficient resolution for our purposes. Clearly the smaller the number entered the finer the resolution, however, there is a limit to the number of contour intervals possible. MacGRIDZO' parameters allow only 256 contour intervals per image, and in some cases the contour interval of 0.25 nT was too small for a contour map to be drawn. Using GridMath, a module included with MacGRIDZO, we were able to filter "grid files" to select a range of z-values to be mapped. For example, by filtering a "grid file" to a maximum z-value of 15 nT we could, in effect, cut off all values above 15, thus allowing a lower contour interval value to be entered and providing better resolution within a particular z-value range of interest. This was particularly important for us in Block 1, where the kilns (discussed below) created a range of over 65 nT, but where, in looking for house walls, we were interested in carefully exploring a range of only about 5 nT. By eliminating the very high and very low readings, we were able to maintain resolution in regions of interest.
While many of the color images we used are included here as
grey-scale images because of the cost involved in reproducting full-color images, it is important to note that the color-filled contour images were the primary ones we used in data analysis and interpretation. We made this choice specifically, basing it on our experiences analyzing color, grey-scale, and dot-density images and our impression that analysis and interpretation were greatly facilitated by the use of color. We were bolstered in this choice by the theoretical literature which suggested our impression was accurate. The logic behind this is simple: humans have evolved color-dependent vision, and we use color as a primary means to distinguish patterns in visual data. Indeed, use of color appears to enhance pattern recognition and recall significantly, a fact taken very seriously by scholars concerned with conveying graphic information.
We used the contour maps produced by MacGRIDZO in a number of ways. We printed them directly as color contour maps, and we also exported them as TIFF and PICT files to the Adobe Photoshop and Canvas software packages. We used Photoshop and Canvas to produce gray-scale images of the MacGRIDZO color-filled conour plots, and to prepare both gray-scale and color images for publication. Canvas allowed us to add other layers of information to the images, for example the Operation 16 excavation plan [cross-reference plan]. All the magnetic images presented in this report were generated using MacGRIDZO and Photoshop.
Results of the Surveys
Our 1993 field work was designed to address three specific issues: (1) the validity of the geomagnetic data for generating maps of archaeological features at Tell es-Sweyhat; (2) the reliability of the instrument and collection method used; and (3) the feasibility of actually mapping the lower town using geomagnetics.
Validity of Geomagnetic Mapping
We defined validity as simply whether the geomagnetic data accurately reflected the actual archaeological deposits. To test the validity of our geomagnetic data we excavated two of our mapped grids. Appendix 4.1, Fig. a shows the archaeological features uncovered in Operation 16 overlaid on the corresponding geomagnetic map. It is clear that the geomagnetic data does accurately reflect the gross features of the archaeological deposits. Specifically, the large magnetic high near the center of the geomagnetic map relates directly to the kiln found in the southwest corner of Operation 16, even suggesting its horseshoe shape. The two smaller kilns in Operation 16 are also obvious on the geomagnetic map. More significantly, the walls forming an angle in the northwest section of Operation 16 can be seen as slight lows on the geomagnetic map, as can the walls branching off from them (except where magnetic highs associated with kilns have obscured them). The walls alongside and behind the large kiln are masked by the kiln's strong magnetic signature.
In addition to Operation 16, the area excavated in Operation 17 was also mapped geomagnetically, but no apparent archaeological anomalies were present. Except for a series of shallow pits, no features or walls were found in the magnetically-flat Operation 17. In both cases, it is clear that the geomagnetic data accurately reflected the presence or absence of archaeological deposits, and was also useful in delineating the nature of archaeological deposits present.
Reliability of Data Collection
We defined reliability as whether geomagnetic data collected at different times or using different collection strategies repeatedly presented the same patterns of subsurface deposits. We tested reliability in two ways. First, we collected geomagnetic data at both 1 m and .5 m intervals in seven separate grids and at different times (we always collected the .5 m interval data first). Appendix 4.1, Fig. b presents .5 m interval data for the area around Operation 16. Appendix 4.1, Fig. c was created using the corresponding 1 m interval data. While there is an obvious decline in resolution, the major magnetic features shown in Appendix 4.1, Fig. a are clearly visible in Appendix 4.1, Fig. c.
Second, we tested reliability by repeatedly gathering data over the same area on subsequent days. A map of geomagnetic data collected over a two-day period and a second map showing the same area collected on a single day several days later and by a different crew were quite similar. Both also show the major archaeological feature of interest, the outer fortification wall, which appears on the maps as a pair of parallel linear magnetic lows, 13-14 m apart. The linear features may represent rampart revetments similar to those uncovered in Operations 25. It is also interesting to us that smaller linear lows branch off from the south of the town wall. We hypothesize that these are the remains of structures built against the town wall similar to those built against the inner town wall in Area IV and the lower town wall in Operation 25.
It is important to note that we did encounter some reliability problems. First, we found that data collected by different individuals did show some differences, but those differences were not enough to mask apparent archaeological features. They were most likely caused by differing heights of the gradiometer above the ground (as some individuals are taller or shorter than others) and differences in the way the gradiometer was held (it should be perfectly vertical, but in practice each individual holds it slighly off-vertical and in a unique way). Second, we found that some items of clothing could effect reliability. Individuals collecting data should not have any metallic objects on their person or clothing, but some metalic items could be easily overlooked. For example, we found that one collector was wearing socks with small metal product tags of which he was totally unaware. We only found them because strange magnetic "streaks" appeared in the data he collected.
What is more important, as the weather got warmer (particulary in June) we found that the instrument became more difficult to "balance" and that the "drift" that occurred while collecting data became greater. Before collecting data the instrument's two sensors must be aligned. This is called balancing, and requires some finesse to do properly, as the instrument is quite sensitive. We found balancing to be more difficult in the afternoons (when the heat was above 100o and the sun was having its greatest daily affect on the earth's magnetic field) than in the mornings (when we balanced the instrument in the cool of dawn). The data for each grid is also calibrated to a zero point. A single reading is taken at that point before collecting data for a given grid, and once the grid is completed a second reading is taken. These readings are used to correct for instrument drift. We found that it was common for the instrument to have zero drift in the morning, but to experience drift of up to 2 nT in the afternoon (a considerable amount when one considers that the walls we located differ from the surrounding soils by only 2 or 3 nT). We believe both are due to heat affecting the instrument's electronics and to the sun's increased distruption of the earth's magnetic field when at its height. However, none of these problems seems to have significantly affected reliability.
Feasibility of Mapping the Lower Town
The main purpose of our 1993 fieldwork was to determine the feasibility of developing a useful archaeological map of the lower town using geomagnetic data. We concluded that the geomagnetic mapping of Tell es-Sweyhat is feasible based on our field experience and the validity and reliability studies we conducted. In only sixteen days of fieldwork (much of it devoted to designing and evaluating different data collection strategies), we were able to map almost 4 ha of the lower town (which because of remapping grids to test reliability represents 2.4 ha or 7% of the lower town's surface). That is far in excess of anything that could be accomplished using standard excavation, even if aided by mechanized techniques. Of more importance, we were able to create useful and informative archaeological maps with the data we collected.
We returned to the site in 1995 to test the results of our 1993 surveys and to continue mapping the lower town. As regards the latter, we decided to focus on collecting data from the northwestern portion of the lower town, an area we were unable to survey in 1993 because it was being irrigated, and from the southern part of the town (south of the wadi that cuts across the lower town) and the extra-mural extension, where we had uncovered tantalizing but severely eroded architectural remains in 1993. As already noted, we chose to collect data in two large blocks over these areas because the one large area we had surveyed in 1993 was much more informative than the smaller blocks. We are still analyzing the 1995 magnetic data.
Archaeological Resources in the Lower Town
We have collected magnetic data fairly broadly across the lower town. While we are only beginning to have enough ground truth from excavations to start the process of actually mapping the archaeological features in these areas, we do have enough information to make at least some preliminary judgements about
the nature and extent of archaeological resources in the lower town. A note of caution is still appropriate. We have only five excavations upon which to base these judgements, and they must necessarily be taken as very tentative until further excavations can provide a firmer basis for interpreting the magnetic data.
Block 1, which encompases nearly 1.5 ha, contains a diversity of archaeological features. Excavations in Operations 16, 23, and 25 have demonstrated the presence of room blocks, pits, and both horsehoe-shaped and circular kilns. The magnetic data suggest room blocks are present across the entire area of Block 1. Additional kilns are also apparent in the area around Operations16 and 23, and we can tentatively conclude that this was a special pottery production site. We have identified what we think may be the inner fortification wall (just visible as a magentic low on the west side of Block 1) and the outer wall (on the east side of Block 1). We have excavated a section of the lower town wall in Operation 25, demonstrating the presence of room blocks built against it as well as earlier structures underlying it. There appears to be a break and offset in the lower town wall due west of Operation 16, and we hypothesize that this might be the location of a gate. A linear feature seems present running from this location to the east. Other curving features branch off it to the north and south just inside (to the west of) the "gate." The linear feature then passes north of Operation 16. We think this feature could be a street running from the city gate through an intersection, where streets branch off it, to the "citadel" Further excavations are clearly needed to establish the identity of the feature.
Block 2 is too small for us to drawn any conclusions about the archaeological deposits present. However, there appears to be a general continuation of the magnetic character of Block 1 into this part of the site, and we tentatively conclude that the region of architectural remains continues here.
The primary archaeological features apparent in Block 3 are those of the lower fortification wall, and the probable walls abutting it within the lower town. These only exist near the wall, and the rest of the block seems devoid of identifiable archaeological features. As discussed above, the excavation in Operation 17 demonstrated that no obvious features are present. If features were present they may have been destroyed by agricultural practices, or this area of the site may have been an open or undeveloped space or a refuse dump. Futher excavation will be needed to establish this interpretation.
Block 4 has not been analyzed as carefully as the other blocks. There appears to be number of magnetic highs that may relate to hearths or ovens, as well as related linear magnetic lows. Block 4 has not been the subject of test excavations either, but we tentatively conclude that architectural features, similar to those uncovered in Operations 4 and 9, may be present in this part of the site.
Block 5 is difficult to interpret. There are several small magnetic highs, similar to those we have been interpreting as hearths, along with some linear magnetic lows, but the two are not as clearly related as we would expect in architectural features (as, for example, are those in Block 4). In addition, these are generally restricted to the northern portion of this block. Wilkinson has suggested that this part of the site has been eroded by a wadi, and perhaps Block 5 crosses the area of erosion, with the northern portion retaining some archaeological debris, but the southern portion retaining none. Without test excavations to refine our ability to discern architectural features, however, we are unable to even tentatively conclude anything about this block.
Blocks 6 and 7 appear basically devoid of archaeological features. There was an isolated, pinpoint magnetic high in Block 6, similar to what we would expect to find associated with a small piece of magnetic metal, but a test excavation in the area of the high did not discover such an object. Block 7 appears to have no significant magnetic features of any kind. We conclude that any archaeological features in this area of the site have been obliterated by agricultural practices or, more likely, by the wadi identified by Wilkinson. Excavations in Operation 19 support this conclusion.
Block 8, covering 3.2 ha, is the largest contiguous block of magnetic data we collected. It appears that surface features such as irrigation channels, ploughing and backdirt piles from earlier excavations, largely dictated the readings obtained. The irrigation channels cut into the area in 1993-94, for example, have created a grid of magnetic lows over the central portion of the block. Despite this, the outer fortification wall is clearly visible in the central and northeastern portion of the block, as are anomalies possibly relating to archaeological features in other parts of the area mapped. As for the outer fortification wall, the Block 8 map appears to show a pair of two parallel linear magnetic lows running from south to north and then curving to the northeast. Low-level aerial photographs show a similar feature, with the outermost (or northern and westernmost) of the pair joining the fortification wall on the north and south. What the innermost (or eastern and southernmost) of the pair represents--perhaps an earlier version of the wall--will have to remain an open question until we can undertake systematic work on the settlement's fortifications. As already noted, Operations 15 and 18 would appear to relate to the innermost of the pair of walls.
In processing collection grids in Block 8, we noted several circular anomalies in the vicinity of the outer fortification wall. Though not directly related to an understanding of the layout of the late third millennium settlement, we suspect these could be tomb chambers. Before undertaking excavations, however, we hope to try other remote sensing techniques to determine if our assumption is correct (see below).
Part of our purpose in collecting data in Block 9 was to see if the wall uncovered in Operation 19 was part of a larger, and perhaps more well-perserved, area of occupation. Although analysis and interpretation of the data from Block 9 has not formally begun, there is a clear network of linear features similar to those interpreted as room blocks in other areas of the lower town. These exist on both sides of the lower town wall, which runs east-west through the center of the block. There are also a number of interesting circular or oval lows, which we are unable to interpret at this time. Finally, there is a large, linear magnetic high at the extreme southern end of the block. The only archaeological feature we have found with similar high readings is the large kiln feature excavated in Operation 23. This, then, may be another kiln. Additional excavations are needed to provide ground truth before interpretation of this block is possible, but it does seem clear that there was activity, and perhaps occupation, in the area south of the lower town wall, and there do appear to be some preserved structures in this extreme southern portion of the lower town.
The magnetic surveys and associated excavations suggest that there is a broad area of significant archaeological deposits in the eastern and northeastern portions of the lower town. These deposits include room blocks and several pottery production sites. The northern section of the lower town appears to have no architectural remains except perhaps abutting the town wall. Other architectural remains in this area may have been eroded or destroyed by agriculture, but this seems improbable given that neigboring sections of the lower town contain architectural remains. This area, therefore, may have been an empty space or perhaps a refuse dump. Excavations suggest that the western portion of the lower town is a second area of significant archaeological deposits, including large room blocks, but the magnetic data have yet to add any insights into this area. Further analysis of the problematic data from Block 8 may provide a better picture of this portion of the site. There appear to be no archaeological deposits of any significance in the south central portion of the lower town. Archaeological remains in this area may well have been eroded away by the wadi that cuts across the site, and we may be unable to reconstruct the nature of this portion of the site. The topographically high area near the southern outer fortification wall, as well as the lower town south may contain significant remains.
This synthesis suggests that future research efforts should be focused on the northern portion of the site and on the extreme southern portion of the lower town (and perhaps the outer town south). Test excavations in surveyed areas should be carried out to provide further information to assist our interpretation of the magnetic data. Excavations should also be carried out to expose clearly delineated houses and the area we hypothesize may be one of the city gates. We hope this work can be initiated in our next field season.
Research during a brief 1996 study season focused on using subsurface interface radar (SIR) and microgravity to map subsurface features in the lower town and, in particular, to located undisturbed tombs in the suspected cemetery at the northwestern edge of the site. As already noted, our geomagnetic mapping of Block 8 showed a number of anomalies that we suspected might represent tombs and we had hoped the additional techniques might confirm or contradict our assumptions. Unfortunately, neither SIR nor microgravity worked to our satisfaction.
The SIR equipment we used was an older Geophysical Survey Systems, Inc. (GSSI) System 8. It is a robust and accurate system, but somewhat cumbersome. The system identifies subsurface interfaces by shooting a broad-band electromagnetic pulse into the ground and graphically recording the returned signal on an electrostatic chart recorder. Soil interfaces and embedded objects are identified by anomalous signal returns, reflected back differentially from the surrounding soil matrix depending on the electrical characteristics of the interfacing soils or objects. For the system to work well, two conditions must be met. First, the soils or objects of interest must have differing electrical characteristics from the surrounding soil matrix. At least in looking for tomb chambers or voids, we knew this condition would be met (air is electrically quite distinct from soil). Second, the soil must have adequate conductivity to allow the electrostatic pulse to penetrate the ground. Most soils are very good insulators in the absence of water and salt and we thought the soils in the area around Tell es-Sweyhat would be appropriate for this technique. We were wrong.
The soils around Tell es-Sweyhat appear to be highly conductive especially at the surface. While one would think this would be good for radar surveying, the opposite is actually true. Conductivity is inversely related to the depth of penetration of an electromagnetic pulse. One can think of soils with high surface conductivity as something like a sponge, soaking up the entire electromagnetic pulse as soon as it hits the ground. Thus, over soils with high surface conductivity it is impossible to shoot enough energy into the soil to generate return signals large enough to measure. And this appears to be precisely the case at Tell es-Sweyhat--we simply couldn't get energy into the soil.
While the SIR proved unusable given the soil conditions around Tell es-Sweyhat, we had good results with the microgravity in locating tombs. We used an extremely sophisticated gravity meter for this job, a Scintrex CG-3 with a sensitivity of less than 5 microgals. The CG-3 measures the intensity of the earth's gravitational field by analyzing the force exerted on a small quartz spring suspending a proof mass within a constant temperature vacuum chamber. When set up perfectly level and allowed to come to rest, the spring will be pulled downward by the proof mass with greater or lesser force in direct proportion to the intensity of the earth's graviational field at that location. The intensity of the earth's gravitational field, in turn, is locally affected by subsurface conditions. Subsurface materials with high mass will locally intensify the earth's gravitational force, while we expected the the tombs' voids to decrease the local intensity of the earth's gravitational field by between 10 and 30 microgals.
Gravity survey is a very slow process, typically taking ten minutes between readings. The gravity meter needs first to be leveled, taking two or three minutes. The quartz spring must then be allowed to stabilize for at least two minutes, and finally a reading can be taken, which requires at least two minutes as the instrument averages a series of instantaneous readings. Given the time-consuming nature of gravity survey, we were only able to collect three data transects: two over a known tomb in the cemetery (Tomb 3), and one over an area we suspected held a tomb because of differential crop growth and a surface depression we noted two years ago (Tomb 4). We had also hoped to collect data transects over suspected tombs identified through geomagnetics, but our very brief season did not allow that. Based on these transects, however, we believe we have identified an undisturbed tomb (Tomb 4).