Appendix: Implementing a Predictive Modelling Program
To formulate a good initial model the ethnohistory of the study' area must be summarized. This provides information on important site settlement modelling criteria, such as average settlement size and population by season, average length of settlement stay by season and site settlement selection by season.
As part of this process, informants from local communities should be sought out whenever possible. Often local informants have a strong understanding of what sites were historically important, where they are located and in what condition they are now in. Using topographic maps and tape recorders, informant data can be collected at specially convened community meetings, band offices and in private homes.
Since there are classes of sites that are difficult to predict because they represent individual rather than group decisions, information about them might be obtained from discussions with First Nations peoples. This process also provides an opportunity to bring the native community into the integrated resource management process.
All known archaeological data should be incorporated into the database. Information from the provincial database can be directly down loaded and checked for errors.
All of these kinds of data are very difficult to use for interpretive purposes because they are not amenable to the inevitable reorganization that is necessary in order to pursue specific resource management problems or answer specific questions. Structuring this data to a usable format involves entering the data into a GIS.
Ecological and ethnological factors form the basis for dividing the study area into smaller units. Each of these subdivisions represents areas where historical cultures had common settlement patterns and lifestyles. This process enables the task of modelling large areas manageable.
Once the site data is compiled, key geographic variables must be derived. The exact techniques used depends on the diversity of the collected data. Usually, the dominant predictive variables become apparent. Lesser trends may require the use of multivariate statistical analysis for clarification.
Derived information is then codified into a set of site location rules. The predictive rules identified in stage one may indicate, for example that certain types of sites are located near streams of a certain size. Based on the best available data, terms such as "near" are statistically quantified, enabling, for example the definition of appropriate buffer zones.
The site prediction rules should be tested periodically against actual site data as it is recovered and the model should be revised accordingly. Also, the priorities and strengths (weighing) of those rules must be re-evaluated periodically. It is likely that numerous changes to the modelling criteria will take place as hypothesis testing is undertaken and better information becomes available.
The classes of geographic data to be used in model building are usually hydrology, drainage, slope, aspect, landforms and vegetation. Much of this information is available from a variety of maps at different scales. other environmental variables must also be considered in order to look for trends which may improve the precision of the model estimates. Some work is usually necessary to structure the existing forms of digital data in order to make it compatible with a mainstream GIS platform.
One hallmark of contemporary attempts at archaeological prediction, and indeed much modern archaeology in general, is the explicit or implicit assumption that environmental factors are major, even exclusive, determinants of much of human behaviour (site location, subsistence strategies, etc.). Environmental variables, such as distance to water, distance to resources assumed to have been important, shelter, and available lookouts, are compared with the location of archaeological materials to determine whether there are correlations between these landscape characteristics and such cultural variables as the location of sites. The causal link between site locations and natural, independent variables is usually considered to be multivariate -that is, people positioned their sites with respect to an optimal combination of all resources in which they were interested.
Environmental factors deemed crucial to site placement usually include vertical distance to water, view, shelter, low slope, and forested resources (Kvamme 1980:96-103). These factors are combined and weighted in a discriminant function analysis. Any given plot of ground can then be rated by this analysis as to its likelihood for containing an archaeological site. This type of model can then be tested against independent data gathered at a later date from within a project area (Klesert 1987:230).
NTS Maps can provide information on topography, altitude range, slope, aspect, and water resources. Geologic maps provide information on the underlying rock formations. Vegetation maps provide information on present vegetation types.
Topographic information may include: crest slope, midslope, bench, foot slope and stream terrace. Two aspects of topography seem important: landform and slope. There appears to be a relationship between water resources and site occurrence, but it is not as clear-cut as for slope or landform. In many cases, sites are located in areas with no water at all, and even the most reliable water resources are not utilized to the same degree as very flat land.
Vegetation is commonly thought to be an important factor for two reasons. First, the vegetation type is important because different types contain different kinds of useful plants in different quantities. Second, borders or ecotones between two types are thought to be especially good locations for sites because they permit use of two vegetation types and therefore a greater variety of useful plants.
However, there is no guarantee that modern vegetation provides an accurate approximation of situations existing thousands of years ago. Much of British Columbia has been subject to logging, mining, grazing and other vegetation-altering activities, and this, in combination with introduced plant species, has altered vegetation patterns.
Finally, it is always important to remember that many sites may have escaped discovery in the course of archaeological survey because of vegetation cover. It is more difficult to discover some types of archaeological sites in some types of vegetation than others. Therefore, the correlation between current vegetation and site location may be due partly to the ease of discovery rather than to prehistoric use patterns.
The representativeness of modern geology is much more certain, but geology has less value for predicting sites. Favourable water resources appear to have increased the toleration for steep slope. Mildly sloping land is often used to its greatest extent near the confluence of streams containing at least one perennial. Landform has the greatest effect, followed by slope, water, and vegetation. The preeminence of landform is natural, since in some categories it combines other factors.
The first testing operations may involve applying regional models against the provincial site inventory. In many areas there may be insufficient number of recorded sites to undertake this kind of test. In areas where this is feasible, testing will be implemented. This is a straightforward process that involves determining the number of sites that are located within or outside the model's predicted areas of high potential or whether sites straddle both high and low potential zones. Testing will also involve making model predictions in areas that are easily accessible (cutovers, road right-of-ways) followed by field assessment of the results. Testing should include some attempt to assess the reliability of informant-collected data by applying model segments which are working well. As more and more data is acquired, the testing cycle will be repeated until an acceptable level of precision and accuracy are obtained.