The existing archaeological site inventory contains a large amount of data which would have utility for predictive modelling. The most useful and, often, the highest quality data occurs in the 21 probabilistic sample surveys from the B.C. Interior. Most of these surveys provide precise details on survey methods and intensity and on areas which can be relied upon to have no sites present. Systematic-intensive surveys of the British Columbia coastline and Interior lakes, rivers, and trails also have some utility. Many lack precise data on surveyed areas where no sites occur, which is the major weakness of the present provincial database.
Standards of survey in British Columbia are controlled adequately for the most part. A major change to the B.C. Archaeological Site Inventory is recommended. It should become a requirement to precisely map actually surveyed areas in reports, and that these plots be transferred to the Inventory maps that show the location of known sites.
The locations of new surveys are not providing data critical to the assessment of archaeological potential in previously unsurveyed areas or environments. Probabilistic surveys no longer seem to be conducted. In the past, these have often been carried out by B.C. Hydro in areas proposed for electrical generation, or by academics for research. Re-survey of some areas, but using sub-surface testing, is recommended to confirm low site potential presently assigned to some low visibility environments. Survey in select unknown areas is recommended in order to acquire the data necessary for predictive modelling. Existing data suggests that the highest archaeological site densities occur in canyons with salmon runs, subalpine areas, and grasslands near water in the ponderosa pine-bunchgrass and Cariboo aspen biogeoclimatic zones. It is estimated that at least 100,000 prehistoric archaeological sites occur in British Columbia.
Predictive modelling and potential mapping should occur at 1:20,000 scale. Potential mapping at 1:250,000 scale should be conducted to alert regional planners and land managers to general areas which have archaeological potential. A recent example of 1:250,000 potential mapping assessed approximately 25% of a very large area as having potential. An older example of potential mapping at 1:50,000 scale within one of these potential areas predicted that 75% of the sites would occur in only 6% of the study area, thus showing the increased power available at the larger scale.
Predictive modelling using logistic regression analysis has been shown to be the most powerful computerized modelling tool in the United States. The normal results obtained through this technique appear to be less accurate than those obtained by the "judgemental" model used to formulate the older potential map described above, which used manual overlays of various environmental variables combined with a probabilistic survey to assign sophisticated potential ratings. Logistic regression should be used to analyse the same data set as that used for the "judgemental" model, and the two maps field tested to determine which is more accurate.
Logistic regression analysis should also be tested to model site distribution using archaeological data gathered during systematic shoreline survey in the Gulf of Georgia combined with a large number of environmental variables, all of which are already stored in OSRIS. The combination of specific environmental variables which best predict archaeological site location are identified during analysis.
