6.0 Advantages of Predictive Modelling

By focusing research on the location of sites, as well as on the types of sites expected to occur in specific locations, the modelling process can help to increase the accuracy and precision of functional, temporal, and spatial qualifiers. Modelling helps to define major similarities and differences among sites and reflects the information potential for both identified and projected sites within an area. In evaluating whether a particular site is potentially significant, the archaeologist often relies on previous experience with other sites of the same type.

The importance of a site cannot be equated solely with its membership in a particular site type or class. However, the rare or unique site, which fails to appear as a separate type during statistical analyses, may be the most significant site in an area. "These sites are often not amenable to identification through sample inventory, but they can be successfully integrated into predictive models if sufficient information is known about them. It is important to consider the physical characteristics of a site as well as its class membership" (Kincaid 1988:562).

Often the nature of the archaeological record itself can indicate that special strategies will be needed for modelling efforts. Sometimes the environment determines whether modelling will be easy or difficult. The most usable environmental variables for predicting site locations are "those that monitor spatial availability of resources (ie. degree of patchiness) and temporal availability of resources (ie. degree of constancy, contingency, and predictability)" (Kincaid 1988:552).

Modelling is useful as a long-term technique for organizing and structuring data and data collection priorities. It is less useful under a short time frame that does not allow for testing and refinement phases.

Perhaps the most valuable applications of modelling, however, is in the area of planning. Planning for the management of cultural resources can take place during the development of land-use plans, environmental assessments, province-wide or area-wide program plans, or site-specific plans. Models are especially suited to planning applications, because they focus on broad scale, generalized trends, actions, or information. The main weakness of models, the inability to consistently produce detailed site-level specific statements, is usually not critical in a regional planning situation (Kincaid 1998:563).

"Are we only interested in predicting general site densities and locations for management purposes, or do we also want to learn about cultural processes and human behaviour? Both of these objectives must be approached from a theoretical standpoint starting with the consideration of how we believe systems of human adaptation operated in the past and evaluating how the ways we discover, collect, and analyze our data are compatible with what we need to know" (Ebert/Kohler 1988:98). The answer may be both, but management decisions are often limited by the amount of money available.

Archaeologists do not know all about the complex systemic behaviour that must be the basis of archaeological predictive modelling. While there may be more than one way to do predictive modelling once archaeologists know how to do it, there is only one way to learn how to do it: by doing it! Developing predictive modelling as a tool to aid both archaeologists and cultural resource managers must proceed from a consideration of the wants and needs of both.

The size of the potential modelling project area is important. Models are most easily interpreted and understood if they relate in a defined way to cultural boundaries or to major environmental zones. When ' only a small portion of a cultural area or environmental zone can be analyzed, only a portion of a cultural system might be examined. In designing small modelling projects, difficulties often occur in meeting minimum sample-size requirements for statistical analyses (Kincaid 1988:551).

The configuration of the modelling project area is also important. Linear as compared to areal projects are generally more difficult to model because linear projects tend to cross-cut several environmental and cultural zones, each of which may be poorly represented regarding the total subject area.

Perhaps the most cost-effective context for model development is within the framework of general planning by a land-managing agency or a local government. These programs (such as the Resource Inventory Committee and the Corporate Resource Inventory Initiative) can develop and sustain long-term approaches that are funded on an incremental basis and result in cumulative and refined data bases. Such data bases, and the models based on them, may take years to develop and test. The end result, however, is a powerful and effective management tool (Kincaid 1988:554).

A common problem encountered by archaeologists is that it can be an extremely difficult task to organize and efficiently manage the large bodies of important data that might exist in a region, let alone the amount of data that occurs on a state/province-wide level. The retrieval of particular kinds of information, for example, about a specific archaeological project, can be exceedingly troublesome when some of the data might exist on one or more maps, while other information might occur on sight forms, project reports, published articles, or even in notes associated with museum collections. The effective value of our recording efforts is thus severely compromised (Kvamme 1989:141).

Since data in GIS have a spatial component, GIS are distinct from traditional database management systems. It is the geographical structure that give GIS added capabilities over traditional database management systems (Kvamme 1989:139).

Maps portraying the geographic arrangement of information are easily produced through a GIS. The visual display of mappable data may be one of the most important features of this technology to the archaeologist, because archaeologists are used to working with maps, and maps convey large quantities of information in an easily understandable and recognizable way. Moreover multiple kinds of information can be displayed simultaneously (Kvamme 1989:162).

Modelling of the regional distribution of archaeological sites is another area of interest to archaeologists and has constituted the largest area of application of GIS technology. "Through GIS a parallel strategy can be employed to yield a regional model of archaeological site location. A variety of terrain, soils, vegetation, hydrological, and other biophysical data, which have been shown by previous work or theory to be related to regional archaeological distributions, are encoded for an area of study within a GIS" (Kvamme 1989:178).

Finally, to evaluate model performance or accuracy, the GIS is used to compare model predictions against test locations of known archaeological class membership. A classification model developed by this strategy can then be regarded as a regional predictive model-when locations of unknown archaeological class membership are classified.

The short-comings of some pre-GIS studies are that they were unable to assess model performance in terms of its application to the entire region of study. For example, "it could be the case that about 97% of all locations (regardless of site presence or absence) in the study area meet 5 of the 7 criteria, the fact that 97% of the known site locations are correctly indicated by the model would not, in this case, be significant. On the other hand, if only 20% of the entire region happened to meet 5 of the 7 criteria, while 97% of the sites did at the same time, then some predictive power would be suggested" (Kvamme 1989:181).

The benefits of archaeological location modelling in GIS contexts are potentially large. Through GIS a wide variety of alternative modelling strategies can rapidly be implemented allowing simultaneous pursuit of a number of different approaches and ideas. Good regional models can help to characterize patterns of prehistoric land use in a more understandable way, often conveying the essence of spatial pattern better than the original site location data. Coupled with their analytic capabilities, GIS potential for advances in understanding of prehistoric cultural processes can be large. Good location models can suggest likely locations where as yet undiscovered archaeological sites might be found. Such models could be extremely useful to archaeologists wishing to discover new sites or to government agencies responsible for the management and protection of archaeological resources on large tracts of public lands.

Parker (1986) argues that GIS-built archaeological models can positively contribute to resource management by increasing the efficiency and quality of management and by reducing costs. Moving a proposed road alignment from a predicted archaeologically sensitive region to an area of low predicted sensitivity, for example, can help to minimize costs for mitigation of impacts. The need for survey and mitigation in these regions is stressed by Parker (1986); however, she emphasizes that the amount of required work would be less because there generally would be fewer cultural resources than in high sensitivity areas (Kvamme 1989:187).