1.1 Introduction
Applications of geographic information systems (GIS) in archaeology may be nearly a decade old, or perhaps a decade and a half, depending on how one defines GIS. Actual use of the term in archaeology begins to appear with regularity around 1983-85 (Hasenstab, 1983a; Kvamme, 1984; Martin and Garrett, 1985; c.f. Pomerantz, 1981), but there are strong antecedents in related technologies that go back to the late 1970s on both sides of the Atlantic.
The earliest literature mostly derives from specialized applications of computer graphics that employed GIS types of operations. Of particular relevance here are studies that utilized spatial databases which produced mapped output. Many pioneering papers also performed various types of manipulations which are clear indicators of GIS, such as weighted layer combinations and the generation of new spatial information. In the following, I focus primarily on North American examples, but also cite parallel developments in Europe.
1.2 The beginnings of GIS in archaeology
1.2.1 Archaeological surface models
The earliest applications akin to GIS clearly lie in computer graphics and statistics. Trend-surface analysis, borrowed from geology, has surprisingly numerous applications from 1975 through the early 1980s (Larson, 1975; Feder, 1979; Hietala and Larson, 1979; Bove, 1981). Typically, polynomial surfaces of various orders were fitted to artifact floors in an attempt to model the distributions of artifacts or other archaeological categories (e.g. lithics or bones). Following naturally from this, some archaeologists explored other types of surface-generating models including weighted-average spatial interpolation methods (Redman and Watson, 1970; Heitala and Larson, 1979; Jermann and Dunnell, 1979). The goal here, again, was to obtain surface generalizing models of artifact distributions to portray their overall patterns of location. This work was greatly facilitated, and in effect promoted, by the wide availability of the first successful spatial analysis and mapping software known as SYMAP (Laboratory for Computer Graphics and Spatial Analysis, 1975), in which we see ancestral forms of GIS functionality including cartographic output, interpolation and generation of surfaces, and manipulation of multiple spatial variables for a single region. It is therefore not surprising that similar applications were occurring in Europe at about the same time (Bradley, 1970; Hodder and Orton, 1976).
1.2.2 Computer mapping and regional databases
Simultaneous with these developments there was early interest in employing computer cartography to generate mapped output for regional spatial databases. One of the best examples is that of Effland (1979) who employed simple pen plotter instructions to provide archaeological site distribution maps for various time periods in the American Southwest. Other contemporary applications employed computer cartography to map artifacts, by type, on excavated occupation floors (Whallon, 1974, 1984; Copp, 1977; Clark, 1979). Here again developments were parallel on both sides of the Atlantic (Flude et al., 1982; Hivernel and Hodder, 1984; Todd et al., 1985), although in Europe there was also an early preoccupation with mapping the results of geophysical surveys (Scollar, 1966, 1974).
1.2.3 Digital elevation models
The application and use of digital elevation models (DEMs) of study region surfaces came quite early in the USA with the work of Scheitlin and Clark (1978), Arnold (1979), Green and Stewart (1983), Kvamme (1983) and others. By the mid-1980s, we see similar work in Britain (McKay, 1984), particularly with the efforts of Harris (1986; 1988). In most of these studies, the DEM was employed solely as a way to visualize patterns better in archaeological and other distributions within a region.
1.2.4 Computer Simulations
The computer simulation work of Zimmerman (1977) in the midwestern USA, and Chadwick (1978; 1979) in Greece, provides another source of GIS precursors. Both employed a variety of environmental data in computer form to examine the problem of human land use and occupation in a dynamic way, through time. Chadwickâs work, in particular, qualifies as an early raster GIS application because he encoded multiple environmental data types in 2 Ă 2 km grid cells and manipulated these to form a weighted composite map of environmental suitability for Early and Middle Helladic Period occupations. This was quite an achievement in the late 1970s, because in terms of functionality it represents much of what we now do with GIS.
1.2.5 The Granite Reef project
The earliest example of a true, full-blown GIS application in archaeology comes from the American Southwest in the Granite Reef archaeological project of 1979-82 (Brown and Rubin, 1982). This project incorporated the services of a professional computer scientist, John Rubin, who (by 1980) had written a complete raster system for processing map data called MAPS (Rubin, 1980). Using this software the granite Reef project established distinct data layers for elevation, soils, geology, rainfall, temperature, and other surfaces over a huge area, approximately 32000 km2, using square grid elements of 3.4 km2. The MAPS program allowed the full manipulation and combination of surfaces, use of map algebra techniques (Tomlin, 1991), and even the derivation of slope, aspect, and other terrain information from the elevation data. These capabilities were fully exploited by the archaeological researchers to develop models, which were weighted layer combinations, of environmental suitability for early hunters, for travel in the desert, and for prehistoric agriculture. These models were then compared against the archaeology to examine goodness-offit.
This project was truly remarkable because, frankly, I see little difference between it and much of what occurs in GIS-based settlement studies today (c.f. Dalla Bona, 1989; Brandt et al., 1992). It is also interesting to note that nowhere in the publication is there any mention of the term âGISâ; the time was simply too early and the term too little known. Brown and Rubin (1982: 272) refer to their operations simply as âa computer-based cartographic analysis systemâ.
1.2.6 Predictive location models
The importance of predictive models of archaeological location to the growth of GIS in North American archaeology cannot be overemphasized. This phenomenon arose principally in the western United States where there exist huge tracts of federally controlled lands. Beginning in the late 1970s, various government agencies began to ask for a means to utilize patterns shown by known archaeological site locations in a region to project or predict where future sites might be located for cultural resource management and planning purposes. Sandra Scholtz-Parker (Scholtz, 1981) in Arkansas, Bob Hasenstab (1983b) in New Jersey, and myself (Kvamme, 1983) in Colorado all independently rose to the challenge and derived amazingly parallel solutions in the form of GIS. This evolution deserves comment.
The basic idea behind the methodology involves the examination of known archaeological sites or settlements in a region for statistical associations with various environmental conditions, such as ground steepness, elevation, aspect, soil, or distance-to-water preferences. Once the statistical associations are found, and multivariate discriminant functions provide an excellent and robust solution, one can go to a map at any point and, based on measurements of the relevant environmental variables, make a decision about the likelihood or even probability of archaeological site presence (Kvamme, 1990a). This procedure could be performed manually at a few locations, but what was needed for management purposes was the systematic mapping, e.g. every 50 m, of the result over large areas to produce archaeological location decision surfaces. In 1980, I hand-measured six variables at 256 contiguous locations spaced 50 m apart (for an 800 Ă 800 m region; Kvamme, 1980), and a year later Scholtz Parker (Scholtz, 1981) did the same for 16 variables, measured at over 7000 locations at a 200 m spacing, to illustrate the potential of archaeological prediction surfaces. Scholtz-Parker, however, manually keyed these data into a statistics-graphics program (the statistical analysis system; SAS Institute, 1978) to derive computer-generated printouts of the hand-measured data and model surfaces. Full automation soon followed. By 1982, I had written computer programs to digitize fully the required map inputs, interpolate DEM, derive analytical surfaces like slope, aspect, and other terrain measures, perform distance operations, and produce predictive model surfaces printed on mylar overlays for a 150 km2 region (Kvamme, 1983); a year later I had undertaken the same for a 1000 km2 region (Kvamme, 1984). Hasenstab (1983b) followed independently with similar results along the Passaic River, New Jersey, using a GIS package written entirely by himself.
During the second half of the 1980s the topic of predictive locational modelling became quite popular (Kohler and Parker, 1986; Judge and Sebastian, 1988). Recent advances in Canada by Dalla Bona (1989) and others are noteworthy, and in the last few years we have seen similar work in The Netherlands (Wansleeben, 1988; Brandt et al., 1992; van Leusen, 1993).
1.2.7 Summary
From the foregoing it should be obvious that if you wanted to use GIS type of methods in the early 1980s you often had to write the programs yourself. Good and reliable software was generally not available at the time. We also se...