GIS AND ARCHAEOLOGY: USING ArcInfo TO INCREASE OUR UNDERSTANDING OF ANCIENT JORDAN

Abstract

Introduction

The value of geographic information systems (GIS) in today's world is well known, but their capabilities also make them ideal tools for the analysis of ancient civilizations. Since 1991, the Madaba Plains Project, in cooperation with the Advanced Resource Technology Group (ART) and the Near East Studies Department at the University of Arizona, has used an ArcInfo based GIS for their archaeological research in the country of Jordan. In many ways, Jordan is an archaeologist's dream, with fantastic monuments, such as Petra, well preserved Roman cities, like Jerash, hundreds of typical Near Eastern tells, innumerable small archaeological sites, and the ubiquitous pottery sherds which are, quite literally, everywhere you look. The Madaba Plains Project has been working in an area just south of Jordan's capital, Amman, since 1968, excavating Tell Hesban, Tell el-Umayri, and Tell Jalul, and carrying out archaeological surveys in the vicinity of each excavation site (LaBianca, 1990; LaBianca, et al., 1995). This paper examines ways in which the Madaba Plains Project has used ArcInfo to further its archaeological research, including the construction of environmentally based site probability models, the use of an erosion model to track the introduction of terrace agriculture during the Iron Age, and the analysis of surface pottery sherds from an excavation site.

Settlement Patterns And Environmental Probability Models

In 1991, the Madaba Plains Project began using ArcInfo to build probability models for sites in the Tell el-Umayri regional survey. Conducted within a five kilometer radius of the main excavation site, the regional survey has documented 131 archaeological sites, ranging in age from the lower Palaeolithic to late Islamic, and in size from urban centers to small, seasonal encampments. The probability models are being used both as predictors of areas likely to produce additional antiquity sites, and to help answer questions about the relationship between archaeological sites and the environment. Already in the early seventies, a distinctive, cyclical pattern was noted in the ceramics collected by both excavation and survey teams on the Madaba Plain (Figure 1); a pattern which is indicative of settlement intensification and abatement in the region. There had always been a suspicion that these cycles were in some way tied to the environment (Geraty & LaBianca, 1985), but making concrete connections between the archaeological sites and the environment had proven difficult. The inclusion of ArcInfo GIS in 1991 finally allowed us to make the connection.

The first step in this process was the construction of an environmental GIS database in GRID. Environmental data, principally hydrography, hypsography, surficial geology, and soils, were digitized from 1:25,000 maps of the Tell el-Umayri region. From these base coverages, environmental variables which detailed everything from simple distance measures, such as distance to wadi channels and soil types, to typical geomorphological themes, such as elevation, slope, aspect, and relief, to more specialized geomorphological themes such as a shelter index, and ridge drainage index, were created in GRID. In all, the Umayri GIS database had 13 environmental coverages.

The second step was to incorporate the archaeological data and build the models, following a methodology developed for archaeological sites by Kvamme (1992). Concentrating here on Iron 1 and Iron 2 sites (a transition from low to high settlement intensity), probability models based on the environmental variables noted above were constructed using stepwise logistic regression. The initial phase in the process was to query all environmental variables using the grid command SAMPLE, in order to determine the local environment at Iron Age I and Iron Age II sites in the region, and, for purposes of comparison, at 250 randomly located non-sites. The results of this process, a series of ASCII files, were exported to STATA (Computing Resource Center, 1992), a PC based statistical software package for analysis. Stepwise logistic regression differs from standard logistic regression in that, based on pre-selected significance levels, it adds and removes variables according to their relative strengths within the regression in order to create a maximum-likelihood model. These significance levels are based on the probability of a given variable's t-score occurring by chance, and for our models aggressive significance levels, 0.15 for adding a variable and 0.20 for removing a variable, were chosen.

Making The Probability Models

Of the thirteen available variables, the stepwise procedure selected four as significant to sites with Iron 1 pottery; Relief Below Locus, Distance to Wadi Channels, Ridge/Drainage Index, and Distance to Type 3 Soils. Figure 2 shows us how these variables were used to create a probability model. To make the model, each variable selected by the stepwise process was multiplied by its corresponding regression coefficient to produce weighted variables. These weighted variables were then summed to create the probability model. Essentially, what this model illustrates is the environmental signature for Iron Age I sites. To clarify the results and facilitate comparisons, the model was logistically transformed, re-scaling its values to a probability score between zero and one for each cell in the grid. The result of this process can be seen in Figure 3a. Here, lighter shades correspond to areas with low probability scores, those areas least like the environmental signature of sites with Iron 1 pottery, and dark shades correspond to areas with high probability scores, those areas most like the environmental signature of these sites.

To test the strength of the model, the grid was queried to determine how well it predicted our site and non-site samples. In the histogram in Figure 4a, there is a clear distinction between the distribution of site and non-site samples, indicative of a fairly strong model. If we use the mid-point in the distribution, 0.5, as a cut-point for predicting site or non-site, we can divide the probability scores into two groups. Looking at the samples to the right of 0.5, we find that the model correctly predicted 81% of the sampled sites, but the price it paid for this level of prediction was that it incorrectly predicted 30% of the non-site sample indicating that this model represents a 51% improvement over chance. In other words, if you were interested in finding more sites with Iron 1 pottery, this model would allow you to concentrate your search in just 30% of the region, with the promise that you will find about 81% of the sites with Iron 1 pottery.

Using the same procedure for sites with Iron 2 pottery, we find a somewhat different picture. Again, the stepwise procedure selected four variables, Maximum Relief, Distance to Type 3 Soil, Ridge/Drainage Index, and Relief Above Locus, to create the model. Already in Figure 3b, we can see that the amount of the survey area falling in the higher probability zones has increased, suggesting that this model is not as strong as the Iron 1 model.

This is supported by the distribution of the site and non-site samples. In Figure 4b, the site sample is still well to the right of 0.5, but now the non-site sample more closely approximates a normal distribution. We can again use 0.5 as a cut-point to divide the region. Now the correctly predicted sites have declined to 65%, while the incorrectly predicted non-sites have risen from 30% to 40%. This leaves us with an improvement over chance of 25%, roughly half of that provided by the Iron 1 model. This is a clear indication that the environmental signature for Iron 2 sites is less specific than that for Iron 1 sites.

What The Probability Models Mean

This brings us to the question of what these models can tell us about the periods they represent. Based on the models explored in this paper, there are clear connections between the environment and locational strategies in the Umayri region. Principally, an inverse relationship between the strength of a model and the level of settlement intensification. As settlement intensifies, the strength of our model decreases, clearly seen in the transition from Iron Age I to Iron Age II. This is evidence of a change in locational strategies. Instead of locating in narrowly focused environmental zones, Iron 2 sites were scattered throughout the environment. This change in the strength of the models suggests both that population pressures were forcing people to settle in less favorable environmental zones, and, as might be expected, a greater diversity in subsistence strategies came with settlement intensification. Finally, extending this logic it could be posited that as available lands become more environmentally marginal, survival pressures would continue to build until a crisis point is reached when environmental and social constraints meet head on with the pressures of population growth, which may explain the sudden, and periodic collapse of settlement in the region.

Erosion Models And Iron Age Agricultural Intensification

Throughout history, soil erosion and water conservation have been important considerations affecting the success of agricultural endeavors. This is especially true in the arid and semiarid Middle East, where intensive agricultural practices, including the construction of agricultural terraces, have been utilized for millennia. Today ancient agricultural terraces litter the landscape, but discovering the age of ancient terrace walls by conventional archaeological means has proven nearly impossible.

This inability to date ancient terrace walls has lead to an ongoing debate concerning the introduction of terrace based agriculture during the Iron Age. Terraces have been portrayed by some researchers as the minimum technological requirement for the establishment of settlements during the Iron Age I period (Stager, 1982; Stager, 1985). Other researchers have argued that the tremendous amount of labor necessary to construct and maintain terraces and the fact that crops can be grown on slopes of up to 30 degrees without the aid of terraces suggests that terraces were not introduced during Iron Age I, a period of relatively low population, but were introduced later as a response to expanding population pressures during Iron Age II (Hopkins, 1985). With no way to ascertain the age of the ancient terraces, the argument remains unsolved.

This project proposes a new approach for addressing the question of when terrace agriculture was introduced Tell el-Umayri. Based on the assumption that if archaeological sites are located in areas with low erosion potential, terracing would have been both unnecessary and improbable, a series of models detailing erosion potential under a variety of conditions were built for the Umayri region. Iron Age I and Iron Age II archaeological sites from the region were then overlaid on the predicted erosion surfaces in order to discover the erosion potential in the vicinity of each site. If it was found that Iron Age I sites were found primarily in low erosion areas, and Iron Age II sites were found in increasingly higher erosion areas, it would indicate that terraces were most likely introduced during Iron Age II.

Constructing the Components of the Erosion Model

This study assumes that areas with high erosion rates (>50 metric tons/hectare/year) cannot be successfully farmed using dryland farming practices without soil and water conservation structures. To model the rate of erosion, the Universal Soil Loss Equation(USLE), developed by the USDA Agricultural Research Service, was used (Wischmeier & Smith, 1978). The USLE is one of the most widely used methods for predicting erosion and has been successfully applied within a GIS framework (Cowen, 1993; Warren et al 1989). One advantage of the USLE is that it uses readily available data which can mapped. The USLE equation is: A = R K L S C, where A, is the computed annual soil loss per acre (metric tons/hectare/year) and is the product of the following factors:

Rainfall Erosivity Factor (R): The R factor reflects the erosive energy of rainfall and is computed based on rainfall intensity. Using information developed by the United Nations Food and Agriculture Organization a R value of 75 was assigned to our study area (FAO, 1979).

Soil Erodibility Factor (K): The K factor is computed as a function of soil texture and structure. The three soils in the study region, referred to as wadi, slope and ridge soils, were all varieties of the Red Mediterranean Soils typical to the region. Wadi soils are found in the drainage bottoms and have a silty clay texture with less than 20% gravel or rock fragments. The slope soils are typically found on slopes and in small tributary wadis. They are a silty clay loam with a gravel or rock fragment content from 30% to 50%. The Ridge Soils, found along the ridge tops, have a silt loam texture and a gravel and rock fragment content between 30% and 70%. Historically, the Ridge soils had the best water and nutrient retention characteristics, and therefore were the better agricultural soils. With the combination of deforestation and discontinuity in terrace maintenance in the region, ridge soils have experienced high erosion levels and today are relatively shallow with significant levels of exposed bedrock. The erosion models assume that this soil was deeper and had a lower percentage of gravel in antiquity than it does today. Using soil survey information from the study area, K values of 0.38, 0.35, and 0.69 were assigned to the Wadi soils, Slope soils, and Ridge soils, respectively. The soil erodibility factor (K) map was created by a reclassification of the soil map of the study area.

Slope Gradient (S) and Length (L) Factors: Contour maps of the region were digitized and a surface model created for the study area. The AML used in the creation of the slope gradient (S) factor was based on a procedure described by Cowen (1993) which uses rasterized facets from a triangulated irregular network, or TIN, to estimate slope gradient. This procedure yielded average slope gradients (in percent) for each facet in the TIN surface which were then rasterized and the S factor computed. Because of resolution limitations in the data a L factor of 1.0 was assumed across the study area.

Vegetative Cover Factor (C): The C factor describes the impact of vegetation, or lack thereof, on potential erosion. Using information from the United States Soil Conservation Service a C factor of 0.01 was assigned to natural conditions, 0.1 to deforested conditions and 0.3 to farmed conditions in the Umayri region.

Building the Erosion Models

Models for three different conditions were constructed; with climax vegetation intact, a deforested landscape, and a deforested landscape being cultivated for agricultural purposes. Creating the models was accomplished by multiplying the USLE factors to create new coverages of erosion potential for the three conditions (Figure 5). These models were then reclassed into three categories based on potential erosion: low erosion potential (<10 metric tons/ha/year), medium erosion potential (10 to 50 metric tons/ha/year), and high erosion potential (>50 metric tons/ha/year). At less than 10 tons, moisture retention is excellent with new soil forming fast enough to replenish that lost to erosion. Between 10 to 50 tons, less water will percolate into the soil and the consequent runoff will cause soil depletion to occur, but at relatively low rates, often slow enough that soil loss would not necessarily be apparent to the farmer. Finally, at rates above 50 tons/ha/year, soil loss would be readily apparent. Sheet wash and gullying of the soil would be problems for the farmer and remedial steps would be necessary for continued farming in these regions.

The predicted erosion with climax vegetation indicate very little potential for erosion (Figure 6a). In this model, 99% of the region loses less than 10 metric tons/ha/year, and the remaining 1% loses between 10 and 50 tons/ha/year. This model demonstrates that left in its natural state, the Umayri survey area would not experience significant erosion problems.

If vegetation were removed from the region, erosion potential changes dramatically, with an increase in areas of high erosion at the expense of the low erosion zones (Figure 6b). The area with less than 10 tons/ha/year of erosion has dropped from 99% to 64%, the area with 10-50 tons/year has risen from 1% to 31%, and 5% of the region with 50 tons/ha/year. Over the long run, deforestation would cause serious problems for the region, and is probably primarily responsible for the large areas of exposed bedrock today.

A more serious change occurs if after the land is cleared it is tilled for agricultural purposes (Figure 6c). This causes the percentage of high erosion zones to increase, with large portions of the area exhibiting high erosion potential. In fact, 24% of the region is now losing more than 50 tons/ha/year, and the areas losing less than 10 tons/year now comprise just 33% of the region. Clearly, the coming of agriculture would have had an impact on the landscape and it was likely a learning process for the farmers as they sought to open new areas to cultivation. It is clear from this model that there are substantial portions of the landscape where terraces would be necessary for agriculture to be successful.

Surface Erosion and Intensification

The erosion models were used to examine the relationship between archaeological sites and erosion potential. Approaching this question from the perspective of the farmer, the expectation was that they would prefer optimal agricultural zones; those with ridge soils and low erosion potential. To obtain an accurate picture of conditions in the immediate vicinity of the sites, Iron 1 and Iron Age II sites were buffered 100 meters, which represents an area of 3.14 hectares per site. The soil and erosion potential maps were then queried by the buffered site maps and cross-tabulations of the results were prepared.

Looking first at the region as a whole, the cross-tabs establish a base from which to make comparisons. In Table 1, erosion potential of less than 50 metric tons/ha/year is labeled acceptable and erosion greater than 50 metric tons/ha/year is labeled unacceptable. Note, that within the region 26% of the land would be considered optimal for agriculture, having both acceptable erosion levels and ridge soils. It is also clear that there was adequate room for the establishment of farms in areas with low potential erosion with only 23.8% of the region falling into the unacceptable erosion category. One final number to note is the total area taken up by ridge soils. At 44%, these soils constitute the best, but not the majority of soil in the region.

Table 1: Cross-tabulation of Soils and Erosion Potential for Survey Area

When compared to the percentages for Iron Age I sites (Table 2) clear differences emerge. As expected a large percentage of the landscape surrounding these sites had both acceptable erosion levels and ridge soils. While this optimal agricultural zone was found in just 26% of the total project area, it constituted 43.8% of the land in the vicinity of Iron Age I sites, making it the largest of the six possible categories. The second thing to notice is that more than 84% of the area surrounding Iron Age I sites is made up of ridge soils, substantially more than the 44% in the project area as a whole. Finally, note the difference between the totals for unacceptable erosion levels. While the region as a whole had only 23.8% in unacceptable zones, the area surrounding Iron Age I sites is 42.2% and all but 1.56% are on ridge soils. Based on these comparisons, the most important factor for Iron Age I settlers was locating in areas with ridge soils. If possible, areas with low erosion were utilized, but the principal factor was clearly soil type.

Table 2: Cross-tabulation of Soils and Erosion Potential for Iron Age I Sites

This conclusion is strengthened if the results for Iron Age II sites are examined (Table 3). Terrace agriculture clearly played a significant role in the Iron Age II economy. If subsistence strategies were significantly different from Iron Age I to Iron Age II, this would be reflected by differences between the samples. Instead, more similarity was found than dissimilarity. There is a shift to the right during the Iron Age II period, as other soil types were utilized more frequently, but the most important factor for Iron Age II sites remains ridge soils, now evenly split between areas of acceptable and areas of unacceptable erosion. Further, the totals for Iron Age II erosion potential are essentially identical to that for the Iron Age I sites, with 42.7% of the area surrounding Iron Age II sites in unacceptable zones. These relatively small differences indicate that the same agricultural zones were being utilized during Iron Age I and Iron Age II. Presumably, they were facing similar problems and utilizing similar strategies to solve these problems.

Table 3: Cross-tabulation of Soils and Erosion Potential for Iron Age II Sites

Erosion Model Summary

GIS was used to examine the relationship between ancient agricultural development and environmental factors. Based on this research several conclusions can be drawn regarding soil erosion and agricultural practices in the Umayri region. First, with deforestation and the introduction of agriculture, soil erosion would have been a serious problem in substantial portions of the region. Second, Iron Age farmers clearly preferred ridge soils. Third, an area's erosion potential was less important than soil type in selecting locations for sites. And finally, it is highly probable that agricultural terraces were utilized by both Iron Age I and Iron Age II farmers. For Iron Age I farmers, the benefit of using ridge soils out weighed the cost of building and maintaining terraces. The terrace system was likely expanded during Iron Age II in response to an increasing population.

Ceramic Distribution And ArcInfo

Finally, during the summer of 1994, the Madaba Plains Project began a long term project using ArcInfo to study surface distribution of pottery sherds at Tell Jalul, Jordan. A scatter of ceramic sherds on the ground is often used by archaeologists to determine the location and time period of an ancient settlement, and they may also reveal something about what an archaeologist would find upon excavating a site. In the Near East, large mounds, or tells, are known settlement sites, often occupied for hundreds or thousands of years, and it has long been held that analysis of the broken bits of pottery scattered across the surface of a tell can help an archaeologist visualize and interpret such things as changing population levels, residence patterns, and social differentiation before excavation begins. Whether or not this assumption is true has never been documented, raising questions about the relationship between surface sherds and still buried archaeological remains.

Data Collection and Processing

In order to begin addressing these questions, it was decided to collect and model the distribution of surface ceramics at Tell Jalul. Excavation at this site was in its earliest stages and took place within the context of a 6 x 6 meter grid of excavation squares, or cells. This same grid was used as the framework for the systematic collection of ceramic and elevation data from the site. There were 45 rows and 55 columns in the grid, creating 2475 cells. Elevation readings were taken from the center point of each cell, and ceramics were collected from every other cell, checkerboard fashion. Each collection cell was referenced by map coordinates at its center point, and assigned a sequential ID number which would serve as a relational item between a point coverage of this grid and the ceramic attribute data tables.

Following shipment to the United States, the ceramics were analyzed and categorized according to their attributes of type -- such as jar, platter, or bowl -- and time period -- such as Early Bronze Age, Iron 1, or Modern. For each cell, the count and weight of ceramics were recorded by vessel type and time period. In all 43,199 sherds, with a total weight of just under 623 kilos, were processed. Of these, 2791 were diagnostic, that is rims, bases, and decorated sherds which can be used to assign a date to the pottery. These ceramic attributes were entered into an INFO table MANYPOTS by cell ID number (ID_NO), with many records per ID_NO. INFO was then used interactively to validate entries, and to calculate total counts (CNT) and weights (WGHT) for all attributes over the entire Tell.

Additionally, the elevation data had been entered into an INFO table. Then a point coverage was created of the center points of each cell in the collection grid with elevation as an item in the PAT. From this point cover, TOPOGRID was used to create a DEM of the Tell, and from this DEM a map of topographic contours was created using LATTICECONTOUR (Figure 7).

Making the Models

Modeling ceramic distribution was a complex process. In INFO, two one-to-many processing methods were used in the production of pottery distribution maps over the tell surface. The first created a "subset" table from selected records. The second method created a summary table containing cumulative totals from items in selected records.

To create the subset table, a selection menu was written in AML; this allowed the user to select an archaeological period which was then passed as a variable to INFO through another AML called INFORESEL. In this AML, the "many" table of ceramic attribute data was the selected table, while an empty template of the subset table was the related table. As you can see, the table (MANYPOTS) was SELECTed (Line 16), then RESELECTed according to the variable value passed from the selection menu (Line 17). The empty table, PSUBSET, was RELATEd with the APPEND option (Line 18).

The next active command created the file by appending one record per relate item to the subset table, automatically filling in the related item's value. In this case, the active command which created the table was the CALC command (Line 19), which transferred the CNT value to the subset CNT item; the next CALC command transferred the WGHT value to the subset table's WGHT item.

Up to this point the INFO commands listed in the AML were processed as if they were executed interactively. This means that a single command processes all selected records. However, through INFO's PROGRAMming feature, it is possible to process several commands per record. The special processing feature of a PROGRAM is the ability to switch between processing one command for all records and processing multiple commands for one record; this switch is in the "odd" and "even" program sections.

The rest of this example demonstrates how a summary table was created, with one record per ID_NO, containing total CNT and total WGHT for the period. In Line 21, the command RUN TOTES ran a program named TOTES in the INFO subdirectory. This program SELECTed the previously created PSUBSET table and RELATEed an empty template table SUMCTWT in the odd program section ONE, where each command acts on all records. Then, in the even program section TWO, both CALC commands acted on the first related record. Processing continued to loop through both CALC commands in the even section (TWO) for each succeeding record until end of table. (If further interactive-type commands were needed, a new odd section THREE would be added at this point.)

The resulting summary-subset table SUMCTWT was used in GRID to create surfaces which model ceramic distribution at Tell Jalul. First, the time period variable was used in ARC to create a coverage containing only points matching the selection criteria. Then SUMCTWT was related by ID_NO to this coverage, and a surface created (Figure 8). The surfaces created by this process show the distribution of pottery from a particular archaeological period, or a particular vessel type across the surface of Tell Jalul. As excavation at the site proceeds, these surfaces will allow us to compare surface ceramic distribution by period, and by vessel type to ceramics found below the surface. This will enable us to answer questions concerning the relationship between surface and subsurface archaeological remains.

Conclusion

The potential of GIS to help archaeologists understand how ancient people related to the landscape is tremendous. Ancient people understood how the land could best meet their needs and located their settlements accordingly. The change in settlement patterns from Iron Age I to Iron Age II in the Umayri region reflects a change in the way they utilized the environment. As evidenced by the strong probability model for Iron Age I sites, a relatively low population allowed these people to focus their settlement in areas best suited to their needs. During Iron Age II, settlement expanded into less desirable areas, reflecting both population pressures and greater diversity in subsistence strategies. This connection between settlement and the environment is also seen in their preference for ridge soils. This soil type was so valued that even in Iron Age I, with its relatively low population, terracing was likely to have been used.

GIS also allows a researcher to see and examine intra-site patterns and make inquiries into how people lived. Although the strengths of GIS lie more naturally with regional studies, the final example shows that it is robust enough to answer site-based questions as well. This kind of flexibility means that the use of GIS, such as ArcInfo, in Near Eastern archaeology will only continue to grow as we use it to examine how ancient cities were organized. As GIS databases expand, inter-regional comparisons will be made and questions regarding topics such as trade and migration addressed. Like everyone in the GIS world, all an archaeologist needs to make this kind of analysis possible is more data, faster computers, improving software, and increased funding.

References Cited

Computing Resource Center (1992). STATA. Santa Monica: Computing Resource Center.

Cowen, J. (1993). A Proposed Method for Calculating the LS Factor for Use With the USLE in a Grid-Based Environment. In Thirteenth Annual Esri User Conference, (pp. 65-74). Palm Springs, CA: Environmental Systems Research Institute.

Food and Agriculture Organization of the United Nations (FAO) (1979). A Provisional Methodology for Soil Degredation Assesment. Rome, Italy: FAO.

Geraty, L. T., & LaBianca, Ø. S. (1985). The Local Environment and Human Food Procuring Strategies in Jordan: The Case of Tell Hesban and its Surrounding Region. In A. Hadidi (Eds.), Studies in the History and Archaeology of Jordan II (pp. 323-330). Amman, Jordan: Department of Antiquities.

Hopkins, D. C. (1985). The Highlands of Canaan. Sheffield, England: Almond.

Kvamme, K. L. (1992). A Predictive Site Location Model on the High Plains: An Example with an Independent Test. Plains Anthropologist, 37(138), 19-50.

LaBianca, Ø. S. (1990). Sedentarization and Nomadization: Food System Cycles at Hesban and Vicinity in Transjordan. Berrien Springs, Michigan: Andrews University/Institute of Archaeology.

LaBianca, Ø. S., Herr, L. G., Younker, R. W., Geraty, L. T., Clark, D. R., Christopherson, G. L., Cole, J. A., Daviau, P. M. M., Fisher, J. R., Lawlor, J. I., Harrison, T. P., Hubbard, L. E., London, G. A., Low, R., & Schnurrenberger, D. (1995). Madaba Plains Project: A Preliminary report on the 1989 Season at Tell el-Umeiri and Vicinity. In W. G. Dever (Eds.), Preliminary Excavation Reports: Sardis, Bir Um Fawakhir, Tell el-Umeiri, The Combined Caesarea Expeditions, and Tell Dothan (pp. 93). Baltimore: American Schools of Oriental Research.

Stager, L. E. (1982). The Archaeology of the East Slope of Jerusalem and the Terraces of the Kidron. Journal of Near Eastern Studies, 41(2), 111-121.

Stager, L. E. (1985). The Archaeology of the Family in Ancient Israel. Bulletin of the American Schools of Oriental Research, 260, 1-35.

Warren, S. D., Diersing, V. E., Thompson, P. J., & Goran, W. D. (1989). An Erosion- Based Land Classification System for Military Installations. Environmental Management, 13, 251-257.

Wischmeier, W. H., & Smith, D. D. (1978). Predicting Rainfall-Erosion Losses - A Guide to Conservation Planning. In USDA Agriculture Handbook No. 282

Author Information

Gary L. Christopherson is a Ph.D. candidate in the Near East Studies Department, and a Research Assistant in the Advanced Resource Technology Group at the University of Arizona 85721. Tel. (520) 621-3045; garych@nexus.srnr.arizona.edu

D. Phillip Guertin is an Associate Professor in the Advanced Resource Technology Group, School of Renewable Natural Resources at the University of Arizona 85721. Tel. (520) 621-1723; phil@nexus.srnr.arizona.edu

Karen A. Borstad is a Ph.D. candidate in the Near East Studies Department, and a Research Assistant in the Advanced Resource Technology Group at the University of Arizona 85721. Tel. (520) 621-3045; borstad@nexus.srnr.arizona.edu