"PREDICTIVE" MODELING OF ARCHAEOLOGICAL DISTRIBUTIONS:
Introductory Concepts

Mathematical and computer models of archaeological location have been around since the early 1980s. They are based on the simple assumption that pattern exists in the places where people locate their activities, camps, or settlements in the landscape. Even today, when camping, we usually select locations that are on level ground, near water, that capture the sun's warmth, that have a good view, that are sheltered from the wind, and so on. So too did ancient peoples. Prehistoric farming settlements may have been located with a concern for good water and soils; bison hunting camps may have been sited with a preference for a view.

Keeping this perspective in mind, there are two ways to go about developing models of where prehistoric peoples located themselves--which translates to where the archaeological sites they left behind might be found. One approach lies in a close examination of the anthropological and historical literature of a region in an effort to deduce the kinds of locations that past peoples may have selected to place their camps and settlements. In this approach the researchers must specify the relevant variables and their values, a difficult task that requires good intuition and educated guesswork, especially when working in the distant past. For example, virtually all peoples tend to place their camps and settlements on level ground, but exactly how level does the ground have to be? Long-term villages might need to be placed near secure water, but how near? Depending on the values chosen different modeling outcomes with widely different accuracies could result.

A far superior approach is to let the people of the past give us clues about the variables that were important to them. This can be accomplished through the technique of archaeological field survey which allows the discovery of archaeological sites in a region through direct field inspection by trained personnel. Simply by measuring and analyzing a large number of environmental variables at known archaeological sites it is possible to ascertain many of the variables that may have had a bearing on past site location decisions. If 25% of a region lies in an unusually productive soil type X, while 75% of known prehistoric farming villages lie in that same soils unit, then we have strong evidence in the data that this variable holds a relationship with the presence of archaeological sites. By examining the data in this way, variable by variable, it is possible to let the past speak to us, to give us clues that allow our model building efforts.

The big practical problem in this task for large regions has always rested in its application. What was needed was a means to map all locations with, for example, level ground, good soils, that were close to water, with south-facing slopes, and so on, according to whatever variables were shown to be important by analysis, over tremendously large regions. It was not until the mid-1980s that this was accomplished using Geographical Information Systems (GIS) computer technology. GIS allows virtually any sort of map variable to be computer-encoded and combined with other variables to yield complex modeling outcomes over very large areas--even across an entire state!

Figure 1. A simple GIS-based model of archaeological site location in southeastern Colorado. The region is approximately 5.5 x 8.5 km in size. (top)

A simplified illustration of a model of archaeological location and associated GIS operations is given in Figure 1. This model, which uses real data from the Great Plains of southeastern Colorado, assumes that the prehistoric occupants preferred low elevation locations, on level ground, and close to permanent sources of water for their camps. Three primary map layers (P) are defined from which all other information, including the model, is derived: (a) elevation, (c) hydrology, and (i) known archaeological sites. A GIS reclassification (R) operation applied to the elevation data (a) yields a low elevation zone (e). A gradient operation (G) on the original elevation surface (a) yields a ground steepness or slope map (b) which is reclassified to portray level ground (f). The initial hydrologic network (c) is reclassified to indicate only permanent water (d), which is subjected to a distance operation (D), yielding a distance to permanent water result (h). This map is then reclassified to form a fixed distance buffer showing locations near water (g). Map layers e, f, and g portray the three criteria assumed to be of relevance to the locations of archaeological sites in this region. We develop the model merely by combining them to find all locations that simultaneously meet all criteria. This is achieved by a GIS operation, known as a boolean intersection (B), with the result shown in j. All that remains is an assessment of how well the model works. This may be undertaken visually, simply by comparing maps i and j, or a GIS cover operation (C) may be employed to drape the known sites over the model result (k) to see how well they match up.

As can be seen, the fit of this model is quite good. More than 90% of the known sites are correctly specified, yet the model includes less than half of the study area. From one point of view a model (j) can be viewed as a description of an archaeological location pattern. From another, it forms a predictive model. The latter occurs when we suggest that a good place to look for new archaeological sites would be in the regions mapped in j. Because the known sites match up so well with the model (k), if other as yet undiscovered sites in the region follow a similar locational pattern, then this model could represent a powerful archaeological prospecting tool.

---

Figure 2. A multiple logistic regression model for the same region in orthographic view. (top)

---

In practice, archaeological model development can be much more complicated, using many more input maps and complex multivariate statistical functions that are optimized to characterize patterns in the data. One such result is shown in Figure 2. It is a multiple logistic regression model based on nine environmental inputs together with a spatial trend model that characterizes very closely the basic location pattern of these data.

Contemporary Modeling Projects (Internet Links)

Mn/Model, sponsored by the Minnesota Department of Transportation, is the largest cultural resource modeling effort ever undertaken, involving thousands of archaeological sites in three culture areas, dozens of variables, 30 m spatial resolution, and approximately 80,000 square miles, spanning the entire state! (KK)

Archaeological Predictive Modeling Program, the Ontario Ministry of Natural Resources, produces models of large size and scope as one aspect of forestry management. (KK)

GIS Modeling of Archaeological Sites in the Raccoon River Greenbelt. The Greenbelt GIS raster database contains data on land cover, soils, elevation, and significant sites (plant, animal, historic, archaeological, geological, and hydrologic). Descriptive modeling was used to compare characteristics of the 85 known sites in the Raccoon River Greenbelt. (KK)

North Carolina GIS Archaeological Predictive Model Project conductucted for the NCDOT by Environmental Services, Inc., and GAI Consultants. (KK)

Grid-Based Modeling for Land Use Planning and Environmental Resource Mapping presented at the 19th ESRI User Conference, by James A. Kuiper. (KK)

Locating Possible Late Pleistocene or Early Holocene Archaeological Sites on the Southern Oregon Coast. (KK)

Predictive modeling the location of archaeological sites by Geonetics Corporation, Boone, NC. (KK)

Settlement patterns modelling through Boolean overlays of social and environmental variables (Islands of Croatia), by Zoran Stancic & Kenneth L.Kvamme. (KK)

Predictive archaeology in Trentino The objective of this study is a map for determining areas of archaeological risk in Trentino (in Italian). (KK)

GIS Techniques, Remote Sensing And Multivariate Models In Assessing Archaeological Resources. (Italy & Nepal) (KK)

Predictive Models of Archaeological Site Distributions in New Zealand. The Pit and Pa predictive model by J.R. Leathwick. (KK)

Archaeological Inventory Study and Archaeological Site Predictive Model: Invermere Pilot. (British Columbia). Greg Anderson, Pilot Leader, 2000. (AS)

A Predictive Model for Archaeological Site Location. (Charleston, SC) John Cable, 1995. (AS)

GIS and Remote Sensing for Archaeology: Bergundy, France. Scott Madry, Informatics International, Inc. (AS)

Archaeology tool finds lost burial site. (Taney County, MO) Marvin Kay, University of Arkansas. (AS)

Agent-Based Modeling of Prehistoric Settlement Systems in the Northern American Southwest. (Soouthwest CO) Timothy A. Kohler (Washington State University) , Carla R. Van West (Statistical Research, Inc.), Eric P. Carr (Carleton College), and Christopher G. Langton (Santa Fe Institute). (AS)

Using GIS to Model and Predict Likely Archaeological Sites. by Christopher Ohm Clement (South Carolina Institute of Anthropology and Archaeology), Sahadeb De (Earth Sciences and Resources Institute, University of South Carolina), Robin Wilson Kloot (Earth Sciences and Resources Institute, University of South Carolina). This is a South Carolina case study utilizing floodplain, hypsographic, and DEM data. (EE)

A GIS Approach for Predicting Prehistoric Site Locations: Upper Chesapeake Bay. by James A. Kuiper and Konnie L. Wescott, This project used environmental data from over 500 known sites in other parts of the Upper Chesapeake Bay region with quite accurate results when compared to the study area. (EE)

Archaeological Predictive Modelling: An Assessment. by Heather Moon, April 30, 1993
This report was submitted to the Resources Inventory Committee (RIC) by the Archaeology Task Group of the Earth Sciences Task Force (Victoria, British Columbia). It provides a very thorough discussion of predictive modeling methods, including a discussion of different types, problems and advantages associated with predictive modeling, and recommendations to the CRM community. (EE)

Predictive Modelling and the Existing Archaeological Inventory in British Columbia. by Morley Eldridge and Alexander Mackie, March 1, 1993. This report was prepared for the Archaeology Task Group of Geology, Soils, and Archaeology Task Force Resources Inventory Committee (Sidney, British Columbia). It describes the quality and coverage of archaeological surveys in British Columbia and the feasibility of incorporating existing data into predictive models. It includes a description of some of the predictive modeling efforts that have been done with the database. (EE)

Locational Modeling of prehistoric site distributions in central Montana. by D. L. Carmichael, 1990. This is Carmichael's web-version of his chapter titled "GIS Predictive Modelling of Prehistoric Site Distributions in Central Montana." In Allen et al. Interpreting Space: GIS and Archaeology, London Taylor & Francis 216-225. The case study is a classic predictive model of prehistoric site distributions located in an 8500 square mile area of north-central Montana. (EE)

The Tides of History: Modeling Native American Use of Recessional Shorelines. by James H. Cleland, Andrew York, and Angela Johnson (KEA Environmental, Inc., San Diego, California). This study creates and implements a 700-year record of changing shorelines of Lake Cahuilla, a huge lake that advanced and receded in response to the changing course of the Colorado River. The rise and fall of the lake level was simulated, allowing the prediction of prehistoric campsites near the shoreline and their movements through time. (EE)

A Rationale for Technical Guidelines on Predictive Locational Modeling of Archaeological Resources on U.S. Army Installations. This is an overview of the objectives and rationale for the development of Department of the Army (DA) guidelines for predictive archaeological modeling in support of the Integrated Cultural Resource Management Plan (ICRMP), which is now required at all Army installations. It includes a discussion of the utility of predictive modeling and the complexity of the methods that need to be understood by resource managers. (EE)

Anthropologist Builds Yucatan GIS to Unlock Secrets of Mayan Settlements. This is a popular article in Earth Observation Magazine (2002) describing Terrance Winemiller's (Department of Geography and Anthropology at LSU in Baton Rouge) settlement analysis and predictive model of Mayan locational behavior. Winemiller is using topography, distance to water, and proximity to limestone, and has plans to use space-based remote sensing images for implementing the model. (EE)

Describing GIS Applications: Spatial Statistics and Weights of Evidence Extension to ArcView in the Analysis of the Distribution of Archaeology Sites in the Landscape. by David T. Hansen (U.S. Bureau of Reclamation, Sacramento, CA.). This site describes the application of the "weights of evidence method" for estimating the distribution of archaeology sites in part of the Central Valley. This method is used to both explore the settlement patterns and to evaluate the predictive model. (EE)

A Bibliography of Predictive Modeling

1. Burton, N. R., and C. A. Shell
   2000 GIS and visualizing the palaeoenvironment. In Computer Application and Quantitative Methods in Archaeology, edited by Kris Lockyear, Timothy J. T. Sly, and Virgil Mihailescu-Birliba, pp. 81-89. BAR International Series 845, Archaeopress, Oxford, England. (EE)
2. Cumming, P.A.B.
   1997 An assessment and enhancement of the sites and monuments record as a predictive tool for cultural resource managment, development control and academic research. Unpublished Ph.D. dissertation, University of Southampton, Southampton.(AS)
3. Custer Jay F., Timothy Eveleigh, Vytautas Klemas, Ian Wells
   1986 Application of LANDSAT Data and Synoptic Remote Sensing to Predictive Models for Prehistoric Archaeological Sites: An Example from the Delaware Coastal Plain. American Antiquity 51(3):572.(AS)
   
4. Dalla Bona, Luke
   1993a Believing is seeing: static and dynamic filters in archaeological research. Manitoba Archaeological Journal 3(1-2): 1-5. (EE)

5. Dalla Bona, Luke 1993b A Preliminary Predictive Model of Prehistoric Activity Location for the Western Lake Nipigon Watershed. Archaeological Computing Newsletter 37: 11-19. (EE)

   Dalla Bona created a simple, yet effective model of the "visual possibility" of archaeological sites dating from 9000 B.P. through the historic period in the Black Sturgeon Lake study area, north of Thunder Bay, Ontario. Using the "value weighted method," five 30-meter-resolution raster layers of environmental variables were used: proximity to water (separated into 6 different subcategories based on permanency), soils, drainages, slope, and aspect. Using weights that apparently seemed appropriate, Dalla Bona created a "visual possibility statement" in the form of a map of the study area with possibilities ranging from 12 to 140. This map was then simplified into areas of low, medium, and high possibility. Known archaeological sites in the study area were then used to evaluate the model, showing that 80% of them occur in the areas identified as high potential, 19.6% in areas of medium potential, and only 0.4% in areas of low potential. The model was then further evaluated by two field seasons of systematic survey and the model was updated to include the new data.

6. Dalla Bona, Luke and Linda Larcombe 1993 Believing is seeing: static and dynamic filters in archaeological research. Manitoba Archaeological Journal 3(1-2): 1-5.

   Dalla Bona and Larcombe describe the problem in Archaeology-particularly in predictive modeling-of static filters. Static filters are at work when scientists conveniently omit certain observations because they do not fit into a preconceived idea about the phenomena of study. In predictive modeling in northern Canada this has led to archaeological site surveys that focus primarily along shorelines-thus filtering out the possibility of finding sites removed from these shores. To avoid this problem, archaeologists need to develop "dynamic filters that structure their thinking and approach to research in such a way that 'other,' 'unknown' and 'aberrant' categories become unnecessary." The authors assert that static filters unconsciously govern archaeological research. To obviate the possibility in predictive modeling, archaeologists need to "turn the blinders off," by trying to establish links between our knowledge of past cultures and the variables that we use to predict the locations of prehistoric activities. This will in turn encourage an explanatory dimension that may not have been previously possible.

7. Dobbs, C.
   1993 Recreating vanished mound groups in the Upper Mississippi River Valley (USA): Integrating Historic Documents, CADD, and Photogrammetric Mapping. In Computing the Past: Computer Applications and Quantitative Methods in Archaeology, edited by Andersen, J. T. Madsen and I. Scollar, pp. 33. Aarhus University Press, Aarhus.
8. Fedje, Daryl W. and Tinea Christensen
   1999 Modeling Paleoshorelines and Locating Early Holocene Coastal Sites in Haida Gwaii. American Antiquity 64(4):635.(AS)
   

9. Gaffney, V., and M. van Leusen 1995 Postcript-GIS, environmental determinism and archaeology: a parallel text. In Archaeology and geographical information systems: a European perspective, edited by Gary Lock and Zoran Stancic, pp. 367-376. Taylor and Francis, Oxford, England. (EE)

   This paper thoroughly presents the main arguments for and against use of predictive modeling in archaeology and the Environmental Determinism (ED) associated with it. The ED that van Leusen refers to is simply the idea that past settlement locations were at least partially determined by the local environmental variables. Several points are made in defense of ED by van Leusen, followed by a rebuttal by Gaffney, and finally some last words by van Leusen. The crux of the debate between the two archaeologists is whether or not ED is a valid assumption. Van Leusen argues that (1) GIS is a tool that was not designed for archaeology, so it is not optimally designed for such a purpose and any shortcomings should be blamed on misuse by the archaeologist rather than the tool itself. (2) ED is justified because (a) it has great utility in CRM work, (b) it makes sense for large temporal scales (since the temporal resolution of archaeological data is often poor), in which social patterns are not detectable but environmental relationships are, and (c) by detecting the patterns in the archaeological record, the non-patterned part of the record can also be determined and extracted. (3) "Cognitive" models are not really that much different than typical (ED) models since they still involve measurable environmental factors. Gaffney replies to all of this that ED is ultimately unproductive because of the tendency with GIS to use whatever data is available regardless of any theoretical relationship. In addition, Gaffney points out the difficulty of representing time-depth in GIS. In general, Gaffney argues that GIS is a trap that all but necessitates environmental determinism, which is an invalid assumption.

10. Gillings, M.
    1995 Flood dynamics and settlement in the Tisza valley of north-east Hungary: GIS and the Upper Tisza project. In Archaeology and Geographical Information Systems: A European Perspective, edited by Lock, G. and Z. Stancic, pp. 67. Taylor and Francis, London.(AS)
11. Goings, C.A.
    2003 A Predictive Model for Lithic Resources in Iowa. Plains Anthropologist 48(184):53. (AS)
12. Gorenflo, L.J. and N. Gale
    1990 Mapping Regional Settlement in Information Space. Journal of Anthropological Archaeology 9:240. (AS)
    

13. Griffin, Dennis, and Thomas E. Chruchill 2000 Cultural resource survey investigations in Kittitas County, Washington: problems relating to the use of a county-wide predictive model and site significance issues. Northwest Anthropological Research Notes 34(2): 137-153. (EE)

    Griffin and Churchill discuss the ill-addressed problems with the Kittitas County Predictive Model developed in 1993 for cultural resource management purposes in Washington State. They lament that the model has been heavily relied upon for land management decisions such as logging for many years while field-testing of the model has shown its many shortcomings. To begin with, the many attempts at model evaluation were well-designed, but quite poorly executed. As a result, the test results precluded any quantitative evaluation, but still qualitatively showed the many shortcomings of the model design and implementation (the subject of this paper). Disappointingly, these problems were ignored, no attempt was made to alter the model, and it was still being used at the time of publication of this paper. The authors suggested the following modifications to the model: a reclassification of the site probability categories, the use of additional criteria to account for the location of historic resources, the addition of oral history interview data, the differentiation between high and low surface visibility in forested areas, and the evaluation of isolated finds with subsurface probing. All of the suggested modifications are based on ample evidence of model failure from the testing procedures mentioned above.

14. Guillot, D. and G. Leroy 1995 The use of GIS for archaeological resource management in France: the SCALA project, with a case study in Picardie. In Archaeology and geographical information systems: a European perspective, edited by Gary Lock and Zoran Stancic, pp. 15-26. Taylor and Francis, Oxford, England. (EE)

    Guillot and Leroy report on a first major attempt to create a functional GIS-based archaeological recording system in France, called the SCALA (a French acronym for an archaeological GIS user interface) project. Though the main goal was to develop a functional GIS for organizational purposes, the authors present a preliminary attempt at predictive modeling at Picardie, a region in north-central France that was undergoing intensive development. The model is very basic and was not complete when the paper was written. The results included the observation of various correlations between prehistoric settlements and physiographic data. However, correlations between settlement density and the landscape were uncertain. No statistical tests were mentioned. Finally, they were able to create a development impact map based on weighting physiographic units according to their observed probability to have archaeological remains at or near the surface.

15. Hammer, John
    1993 New predictive site location model for interior New York State. Man in the Northeast 45: 39-76. (EE)
16. Hasenstab, Robert
    1991 Wetlands as a critical variable in predictive modeling of prehistoric site locations: a case study from the Passaic River basin. Man in the Northeast 42: 39-61. (EE)
17. Kamermans, H. and M. Wansleeben
    1999 Predictive Modelling in Dutch Archaeology, Joining Forces. BAR International Series 757 (AS)
18. Kirkinen, T.
    1996 Use of Geographical Information System (GIS) in Modeling the Late Iron Age Settlement in Eastern Finland. Environmental Studies in Eastern Finland - Helsinki (8):19-61. (AS)
    
19. Krist Jr., Frank J, and Danial G. Brown
    1994 GIS Modeling of Paleo-Indian Period Caribou Migrations and Viewsheds in Northeastern Lower Michigan. Photogrammetric Engineering and Remote Sensing 60(9): 1129-1137. (EE)
20. Lake, M.W.
    2000 Computer simulation of Mesolithic foraging. In Dynamics in human and primate societies: agent-based modelling of social and spatial processes, edited by Gumerman, G. and T.A. Kohler, Oxford University Press, Oxford. (AS)
21. Lock, G. (ed) 1999 Beyond the Map: Archaeology and Spatial Technologies. NATO Science Series A: Life Sciences, Amsterdam: IOS Press. (AS)
    

22. Lookabill, Anna B. 1998 Predictive model for locating vaccinium-huckleberry processing sites in the northern Cascades of Washington. Northwest Anthropological Research Notes 32(2): 173-181. (EE)

    Lookabill suggests that archaeologists in the Pacific Northwest have largely ignored the utilization of plants by prehistoric peoples and as a result base predictive models on hunting and fishing. Recently, however, a number of Huckleberry processing sites have been discovered in the southern Washington Cascade Mountains area. Since the northern Cascade Mountains area is similar to the southern portion with regard to the environment and the archaeological record, it is likely that similar sites may exist in the north. The author proposes a predictive model of Huckleberry processing sites in the northern Cascade Mountains of Washington state by utilizing associations between the known sites and environmental variables in the southern Cascade Mountains area. However, due to changing environmental conditions and the nature of the Huckleberry plant, prediction is difficult and Lookabill identifies only elevation as having predictive possibilities. She proposes to isolate the significant elevation range and perform a random sampling in that area in search of Huckleberry processing sites. She further suggests that since Huckleberry processing was a female activity, this would be one step in the right direction toward a more holistic (less male-centered) view of past cultures.

23. Loots, G.L.
    The use of projective and reflective viewsheds in the analysis of the Hellenistic defence system at Sagalassos. Archaeological Computing Newsletter 4912. (AS)
    

24. Lopata, Elena Parent, and Shih-Lung Shaw 1992 Searching for Sunken Treasure in Turkey's Sea of Marmara. Geo Info Systems 2 (7): 57-61. (EE)

    Lopata et al. made a simple predictive model for the location of well-preserved shipwrecks in the Sea of Marmara-the only waterway between the Black and Aegean seas. Using bathymetry data (at a 1: 271,690 scale), colony locations, marble quarry sites, and estimates of the most likely trade routes (all vector layers), the likelihood of finding sunken ships from Hittites, Assyrians, Myceneans and Greeks was calculated. The authors chose water depths of greater than 100 meters as the most likely to preserve shipwrecks because of the presumed anoxic conditions. In addition, they isolated areas within 3000 meters of shorelines (including islands and large rocks). The intersection of these two binary layers produced a large "search area" for shipwrecks. This area was then narrowed by creating a "probability grid," which was an additive model of a series of buffers around ancient ports and quarries, navigable straights, and shipping routes. The resulting coverage ranged from 0 to 16 in likelihood, and was categorized into four categories. The highest potential category was chosen as the target search area. As a qualitative evaluation of the results, the authors note that an area that was previously designated as having high potential for shipwrecks by a well-known researcher was agreeable to these results. No empirical data were available at the time for further evaluation.

25. Maschner, H.D.G. (ed)
    1996 New methods, old problems: geographical information systems in modern archaeological research. Carbondale: Center for Archaeological Investigations, Southern Illinois University at Carbondale. (AS)
    
26. Maschner, Herbert D. G. and Jeffrey W. Stein
    1995 Multivariate approaches to site location on the Northwest Coast of North America. Antiquity 69: 61-73. (EE)
    

27. Perkins, Philip 2000 A GIS investigation of site location and landscape relationships in the Albegna Valley, Tuscany. In Computer Application and Quantitative Methods in Archaeology, edited by Kris Lockyear, Timothy J. T. Sly, and Virgil Mihailescu-Birliba, pp. 133-140. BAR International Series 845, Archaeopress, Oxford, England. (EE)

    Perkins states that site location may be thought of in two ways: "from the viewpoint of an individual site in the landscape or from the viewpoint of the landscape which is partially occupied by sites." (p. 133) In his analysis of the settlement pattern in the Etruscan period in Tuscany, he employs the latter viewpoint because it is though to be more applicable to complex societies. To do this he uses a chi squared test to show statistically significant associations between settlement locations and categories of altitude, slope, aspect and solid geology. He then uses Yule's Q to determine if the associations are positive or negative, and evaluates the strength of the associations with a f2 test. Ten different time periods (centuries) were explored in this way, allowing the changes through time to be determined. For each time period and for both positive and negative associations, additive models were created with index values ranging from 1 to 4. The time slices were then lumped into three periods for ease of analysis. The results showed varying associations between sites and the variables through time, with a pronounced shift in settlement locations after the Roman conquest.

28. Phillips, John C., and Steve Duncan 1993 Archaeology and the geographic resources analysis support system: a preliminary model of archaeological site location in Santa Rosa County, Florida. Florida Anthropologist 46(4): 251-262. (EE)

    Phillips and Duncan created a correlative predictive model of archaeological sites based on known site locations and their correlations with one environmental variable: soil type. This was thought to be a fair indicator based on the idea that soil types are defined by other variables such as "slope, drainage, property, texture, composition, and vegetation." Thus soil types are markers of specific environmental zones. Due to the long life or soils and their slow rate of change, the authors believed that soil type was better than more ephemeral variables such as water sources and vegetation. The type of soil data is presumably a vector coverage, though the details are not mentioned in the text. It is also presumably high in resolution, since soil type is broken up into 50 categories. The resulting model simply shows those soil polygons with high "probabilities" of archaeological deposits. It was evaluated by a systematic, but limited field survey entailing 10 transects for a total of 16.7 linear miles, covering both high and low probability zones. The surveys included visual reconnaissance of the surface as well as shovel test pits every 30 meters in areas of low visibility. The new data showed that the model was quite good for predicting sites in the coastal lowlands. However, for the interior Western Highlands portion of the study area sites were found in areas of zero probability while the areas of high probability yielded no sites.

29. Wheatley, David
    1996 Between the lines: the role of GIS-based predictive modeling in the interpretation of extensive survey data. In Interfacing the past: computer applications and quantitative methods in archaeology CAA95 vol. II, edited by Hans Kamermans and Kelly Fennema, pp. 275-292. University of Leiden, Leiden, The Netherlands. (EE)
    
30. Wheatley, David.
    2000 Spatial technology and archaeological theory revisited. In Computer Application and Quantitative Methods in Archaeology, edited by Kris Lockyear, Timothy J. T. Sly, and Virgil Mihailescu-Birliba, pp. 123-131. BAR International Series 845, Archaeopress, Oxford, England. (EE)
    
31. Wise, Alicia L.
    2000 Building theory into GIS-based landscape analysis. In Computer Application and Quantitative Methods in Archaeology, edited by Kris Lockyear, Timothy J. T. Sly, and Virgil Mihailescu-Birliba, pp. 141-147. BAR International Series 845, Archaeopress, Oxford, England. (EE)

32. Woodman, Patricia E., and Mark Woodward. 2002 The use and abuse of statistical methods in archaeological site location modelling. In Contemporary Themes in Archaeological Computing, edited by David Wheatley, Graeme Earl and Sarah Poppy, pp. 22-27. University of Southampton Department of Archaeology Monograph No. 3. Oxbow Books, Oxford, England.

    Woodman and Woodman (2002) describe some of the problems with statistical methods used in inductive site location models that have not yet been addressed. They state that there are three main assumptions for statistical tests that have been ignored in archaeological predictive modeling: data collection/sampling integrity, a well established relationship between the sites and the variable used to predict their location, and complete independence of each variable including the uniqueness of its affect on site location. In regard to the second problem, assumption of a linear relationship between a variable and site location is often ignored in logistical regression. For example, elevation often does not have a linear relationship with the probability of settlement location yet it is often treated that way. The authors call for greater attention to establishing the relationship between site location and the variables used to predict them, and the development of appropriate statistical tests for linearity. Finally, there is the problem of the relationship between the variables used for the models. Even if they are shown to be completely independent from one another, they can still have a "confounding" and/or an "interaction" relationship with one another. "Confounding" refers to the problem that two variables, while uncorrelated, may actually substitute for one another in their ability to predict site location. "Interaction" occurs when two variables in combination make one good predictor whereas alone they are weak. The authors do not offer any solutions to these problems.

Predictive Modeling References by Kenneth L. Kvamme

Publications (Principal works in topic are highlighted)

1. (1980) Predictive model of site location in the Glenwood Springs Resource Area (Chapter 7). In A Class II Cultural Resource Inventory of the Bureau of Land Management's Glenwood Springs Resource Area. Report submitted to Bureau of Land Management, Grand Junction District, Colorado by Nickens and Associates, Montrose, Colorado.
2. (1983) New Methods for Investigating the Environmental Basis of Prehistoric Site Locations. Ph.D. dissertation, Department of Anthropology, University of California at Santa Barbara.
3. (1983) A Manual for Predictive Site Location Models: Examples from the Grand Junction District, Colorado. Ms. prepared for Bureau of Land Management, Grand Junction District, Colorado.
4. (1983) Computer Processing Techniques for Regional Modeling of Archaeological Site Locations. Advances in Computer Archaeology 1:26-52.
5. (1984) Models of Prehistoric Site Location Near Pinon Canyon, Colorado. In Papers of the Philmont Conference on the Archaeology of Northeastern New Mexico, C. J. Condie, ed., New Mexico Archaeological Council Proceedings 6, pp.349-370.
6. (1986) The Use of Geographic Information Systems for Modeling Archaeological Site Distributions. In Geographic Information Systems in Government (Vol.1), B. K. Opitz, ed., A. Deepak Publishing, Hampton, Virginia, pp.345-362.
7. (1988) Using Existing Archaeological Survey Data For Model Building. In Quantifying the Present and Predicting the Past: Theory, Method, and Application of Archaeological Predictive Modeling, W.J. Judge and L. Sebastian, eds., U.S. Department of the Interior, Bureau of Land Management, U.S. Government Printing Office, Washington, D.C., pp. 301-323.
8. (1988) Development and Testing of Quantitative Models. In Quantifying the Present and Predicting the Past: Theory, Method, and Application of Archaeological Predictive Modeling, W. J. Judge and L. Sebastian eds., U.S. Government Printing Office, Washington, D.C., pp.325-428.
9. Kvamme, K. L., and T. A. Kohler (1988) Geographic Information Systems: Technical Aids for Data Collection, Analysis, and Display. In Quantifying the Present and Predicting the Past: Theory, Method, and Application of Archaeological Predictive Modeling, W. J. Judge and L. Sebastian eds., U.S. Government Printing Office, Washington, D.C., pp. 493-547.
10. K. L. Kvamme and M.A Jochim (1989) The Environmental Basis of Mesolithic Settlement. In The Mesolithic in Europe: Papers Presented at the Third International Symposium, C. Bonsall, ed., John Donald Publishers, Ltd., Edinburgh, pp. 1-12.
11. (1989) Geographic Information Systems in Regional Archaeological Research and Data Management. Archaeological Method and Theory, Vol. 1, M.B. Schiffer, ed., University of Arizona Press, Tucson, pp. 139-202.
12. (1990) GIS Algorithms and their Effects on Regional Archaeological Analysis. In Interpreting Space: GIS and Archaeology, K.M.S. Allen, S.W. Green, and E.B.W. Zubrow, eds., Taylor and Francis, Ltd., London, pp. 112-126.
13. (1990) Predictive Cultural Resource Modeling at Pinon Canyon: 1984-85. In An Introduction to the Archaeology of Pinon Canyon, Southeastern Colorado, Volume I: Background and Methods, W. Andrefsky, ed., National Park Service, Rocky Mountain Regional Office, Denver, Colorado, pp. (IV:)6-83.
14. (1990) The Fundamental Principles and Practice of Predictive Archaeological Modeling. In Mathematics and Information Science in Archaeology: A Flexible Framework. A. Voorrips ed., Studies in Modern Archaeology, Vol. 3, Holos-Verlag, Bonn, Germany, pp. 257-295.
15. (1992) Geographic Information Systems and Archaeology. In Computer Applications and Quantitative Methods in Archaeology 1991, G. Lock and J. Moffett, eds., BAR International Series S577, Tempus Reparatum, Oxford, pp. 77-84.
16. (1992) A Predictive Site Location Model on the High Plains: An Example with an Independent Test. Plains Anthropologist 37: 19-40.
17. R. Brandt, B. J. Groenewoudt, and K. L. Kvamme (1992) An Experiment in Archaeological Site Location: Modeling in the Netherlands Using GIS Techniques. World Archaeology 24: 268-282.
18. (1997) Modeling Archaeological Distributions. Mn/Model Newsletter (Minnesota Department of Transportation) 3(1):2- 3.
19. (1999) Recent Directions and Developments in Geographical Information Systems. Journal of Archaeological Research, 7(2):153-201.
20. Z. Stancic and K.L. Kvamme (1999) Settlement Pattern Modelling Through Boolean Overlays of Social and Environmental Variables. In New Techniques for Old Times, CAA98, J.A. Barcelo, I. Briz, and A. Vila, eds., BAR International Series 757, Tempus Reparatum, Oxford, pp. 231-237.

(top)

Conference Presentations & Invited Lectures

1. "Alternative Methodologies for Settlement Pattern Analysis: The Determination of Factors Regulating Prehistoric Site Placement and Site Location Predictions," Society for American Archaeology meeting, San Diego (1981).
2. "Understanding and Explaining Hunter-Gatherer Site Locations as a Function of Environmental Variation," Society for American Archaeology meeting, Minneapolis (1982).
3. "A Computer Approach for Wide-Area Probability Mapping of Archaeological Site Presence," Society for American Archaeology meeting, Pittsburgh (1983).
4. "Predicting/Modeling Archaeological Site Locations: An Overview of Techniques and Problems in the Grand Junction District, Colorado," Plains Anthropological Conference, Rapid City, South Dakota (1983).
5. "Analyzing and Modeling Prehistoric Settlement Patterns in Central Colorado." Department of Anthropology, University of Montana, Missoula (1983).
6. "Modeling Archaeological Site Locations at Piņon Canyon, Colorado," Colorado Archaeological Society annual meeting, Boulder (1984).
7. "New Techniques for the Investigation of Archaeological Settlement Data." Department of Anthropology, University of California, Santa Barbara (1984).
8. "Settlement Studies at Piņon Canyon, Colorado," Northeast New Mexico Archaeological Conference, Philmont Scout Ranch (1984).
9. "Pattern Classification Techniques and Geographic Information Systems for Prehistoric Site Location Modeling." Department of Anthropology, University of Pennsylvania, Philadelphia (1985).
10. "Pattern Classification Techniques and Geographic Information Systems for Prehistoric Site Location Modeling." Arizona State Museum, University of Arizona (1985).
11. "Geographic Information Systems Techniques for Regional Archaeological Research," International Symposium on Data Management and Mathematical Methods in Archaeology, International Union of Pre- and Protohistoric Sciences (Commission IV), Denver, Colorado (1985).
12. "Fundamentals and Potential of Geographic Information Systems Techniques or Archaeological Spatial Research," presented in the symposium Computer-Based Geographic Information Systems: A Tool of the Future for Solving Problems in the Past, organized by K.L. Kvamme and R.J. Hasenstab, Society for American Archaeology meeting, Denver (1985).
13. "The Use of GIS for Modeling Archaeological Site Distributions," presented in the Geographic Systems in Government Workshop, sponsored by the U.S. Army Engineer Topographical Laboratories, Washington, D.C (1985).
14. "The Environmental Basis of Mesolithic Settlement," co-author: M.A. Jochim, Third International Symposium on the Mesolithic in Europe, International Union of Pre- and Protohistoric Sciences, Edinburgh, Scotland (1985).
15. "Archaeological Distributions, Geographic Information Systems, and Problem-Oriented Research in Archaeology." Albert C. Spaulding Distinguished Lecturer Series, Institute for Quantitative Archaeology, Department of Anthropology, University of Arkansas, Fayetteville (1986).
16. "Overview of Geographic Information Systems for Archaeological Research and Data Management," presented in the National Workshop on Microcomputers in Archaeology, sponsored by the Society for American Archaeology and the Society of Professional Archaeologists, New Orleans (1986).
17. "GIS Algorithms and Data Quality: Their Effects on Spatial Models," Arizona Conference on Remote Sensing and Geographic Information Systems in Management, sponsored by the Remote Sensing Center, Office of Arid Lands, University of Arizona, Tucson (1988).
18. "GIS Algorithms, Analysis Procedure, and Regional Archaeological Spatial Analysis," presented
19. in the symposium Geographic Information Systems Applications in Archaeology, Society for American Archaeology meeting, Phoenix (1988).
20. "Archaeology and GIS in the 21st Century," poster at the Society for American Archaeology meeting, Las Vegas (1990).
21. "Archaeological Site Prediction," presented in the Arizona GIS Workshop: Geographical Information Systems for Personal Computers, sponsored by the U.S. National Park Service at the University of Arizona, Tucson (1991).
22. "Spatial-Locational Analysis Through GIS," Computer Applications and Quantitative Methods in Archaeology Conference (CAA), Oxford, England (1991).
23. "Regional Analysis and Modeling with GIS." Albert Egges van Giffen Institute for Pre- and Protohistoric Archaeology, University of Amsterdam, Netherlands (1991).
24. "An Overview of GIS Applications in Archaeology." ROB (Rijkdienst voor het Oudheidkundig Bodemonderzoek) - State Archaeological Service, Amersfoort, Netherlands (1991).
25. "Some Applications of GIS in Regional Archaeological Studies." Department of Anthropology, Ohio State University, Columbus (1991).
26. "Spatial-Locational Analysis and Modeling with GIS." Archbold Biological Station, Lake Placid, Florida (1991).
27. "Archaeology and Geographic Information Systems." Arizona State Land Department Colloquium Series, Phoenix (1992).
28. "GIS for Archaeological Survey and Planning." Department of Archaeology, University of Ljubljana, Slovenia (1992).
29. "Analyzing and Modeling the Environmental Aspect of Prehistoric Distributions with GIS." Institute of Archaeology, Oxford University, England (1992).
30. "GIS Applications in Archaeological Analysis and Modeling." Department of Archaeology, Cambridge University, England (1992).
31. "GIS and Archaeology." Department of Archaeology, University of Leicester, England (1992).
32. "The Application of Geographic Information Systems in Archaeology: New Developments." Center of Oriental Archaeology, Paris I University, France (1992).
33. "GIS in Archaeology: Some Recent Examples." Biological-Archaeological Institute, University of Groningen, Netherlands (1992).
34. "The History and Practice of Predictive Modelling in Archaeology," presented in the Symposium on Archaeological Predictive Models at the Canadian Archaeological Association (CAA) Conference, Montreal (1993).
35. "Theory, Method, and GIS in Landscape Archaeology," presented in the symposium Landscape Archaeology: The New Integrative Synthesis at the Plains Anthropological Conference, Saskatoon, Saskatchewan, Canada (1993).
36. "GIS and Imaging Techniques as a Foundation for Regional Archaeological Research." Department of Archaeology, Boston University (1993).
37. "GIS Techniques as a Foundation for Regional Archaeological Research." Desert Archaeology, Inc., Tucson, Arizona (1993).
38. "A View From Across the Water: The North American Experience in Archaeological GIS." Presented in the international conference The Impact of Geographic Information Systems in Archaeology: A European Perspective, European University Center for Cultural Heritage, Ravello, Italy (1993).
39. "GIS-Based Procedures for Modeling and Predicting Prehistoric Settlement Locations." Presented in the Computers in Academia lecture series, Computing and Information Services, University of Pittsburgh (1994).
40. "GIS-Based Procedures for Modeling Prehistoric Settlement Distributions." Presented in the Center for Computational Science Seminar, Center for Computational Science, Boston University (1994).
41. "Geographic Information Systems: Computer Tools of the Present for Solving Problems of the Past," Massachusetts Archaeological Society annual meeting, Middleboro (1995).
42. "Regional Spatial Analysis Using GIS," Presented at the Center for Scientific Research of the Slovene Academy of Sciences and Arts, Spatial Information Center, University of Ljubljana, Slovenia (1995).
43. "Archaeological Spatial Analysis Using GIS: Methods and Issues," Presented in the course Information Systems and Geographical Networks in Archaeology: GIS and INTERNET, organized by R. Francovich, A. Gottarelli and D. Manacorda, sponsored by the Department of Archaeology, University of Sienna and the Italian Council of Research, Sienna, Italy (1995).
44. "Archaeological Applications of IDRISI." Presented at the IDRISI User's Conference, Clark University, Worcester, Massachusetts (1996).
45. "Analyzing Regional Archaeological Site Location Patterns: Combining Quantitative and Qualitative Approaches with GIS." Presented at the Department of Anthropology, Rutgers University (1996).
46. "Regional Studies in the Southwest and Beyond: Small and Large Scale Approaches through GIS Technology." Presented at the Department of Anthropology, Northern Arizona University, Flagstaff (1997).
47. "GIS Applications in Archaeology."  Presented at the Museum of Anthropology, University of Michigan, Ann Arbor (1997).
48. "Evaluating Models: Several Examples from the United States."  Presented at the workshop Addressing Archaeological Predictive Modelling in the Boreal Forest, sponsored by the Ontario Ministry of Natural Resources Land Use Planning Branch, Sault Ste. Marie, Ontario (1997).
49. "Statistical Location Analysis in Archaeological Predictive Models." Presented at the Department of Archaeology and Anthropology, University of Saskatchewan, Saskatoon (1998).
50. "Analyzing and Modeling Settlement Patterns Through Boolean Overlays of Social and Environmental Variables." Co-author: Zoran Stancic, Computer Applications and Quantitative Methods in Archaeology Conference (CAA), Barcelona, Spain (1998).
51. "There and Back Again: Views on the Development of Predictive Location Modeling in Archaeology," banquet presentation for the conference: GIS and Archaeological Predictive Modeling: Large-scale Approaches to Establish a Baseline for Site Location Models, Chicago (2001).
    (top)

BACK TO PROJECTS

(last updated: 2-25-03)